JP3580168B2

JP3580168B2 - Loudspeaker

Info

Publication number: JP3580168B2
Application number: JP07292799A
Authority: JP
Inventors: 実福島; 博昭竹山; 章寺澤
Original assignee: Matsushita Electric Works Ltd
Current assignee: Panasonic Electric Works Co Ltd
Priority date: 1999-03-18
Filing date: 1999-03-18
Publication date: 2004-10-20
Anticipated expiration: 2019-03-18
Also published as: JP2000270089A

Description

【０００１】
【発明の属する技術分野】
本発明は、家庭内、ビルディング、工場等で用いられる拡声通話装置に関するものである。
【０００２】
【従来の技術】
従来より、インターホンや電話機あるいはＰＨＳ等の拡声通話装置においては、スピーカからマイクロホンへの音響フィードバックおよびハイブリッド回路（２−４線変換回路）におけるインピーダンスの不整合により閉ループが形成され、増幅器の利得が大きすぎる等の理由により上記閉ループの利得が１倍以上になるとハウリングが生じるという問題があり、ハウリングが生じた場合には通話を継続することが困難となるため、ハウリングを抑圧する手段が必要不可欠となっていた。
【０００３】
そこで従来は、送受話信号を監視することで通話状態が受話状態か送話状態の何れであるかを判別し、受話状態の時には送話信号に対して所定の損失量を挿入し、送話状態の時は受話信号に対して所定の損失量を挿入することにより、閉ループの一巡伝達利得を低減させる、所謂音声スイッチと呼ばれるものが広く用いられてきた。この音声スイッチを用いる場合、挿入損失量を大きくしすぎると通話中に切断感を生じる等の通話品質の劣化を招くため、挿入損失量は所望の利得余裕（ハウリングマージン）が得られるだけの必要最小限とすることが望ましい。
【０００４】
上述のような課題に対して、近端側（音声スイッチの挿入位置よりも端末側）と遠端側（音声スイッチの挿入位置よりも回線側）の帰還利得を随時推定し、その推定結果に応じて挿入損失量を適応的に制御することによって、常に一定の利得余裕を保ちながら挿入損失量を必要最小限とする音声スイッチが提供されている（特開昭６３−２１２２５１号公報等参照）。
【０００５】
上記公報に記載されている従来技術の原理を図６を参照して説明する。近端側のマイクロホン１に送話減衰器３０が接続されるとともにスピーカ３に受話減衰器３１が接続されているとき、受話減衰器３１の入出力信号ｒ_ｉｎ，ｒ_ｏｕｔを整流平滑して求められる振幅値をＲｉ，Ｒｏ、送話減衰器３０の入出力信号ｓ_ｉｎ，ｓ_ｏｕｔを整流平滑して求められる振幅値をＳｉ，Ｓｏと定義する。そして、近端側の帰還利得αを推定するために送話減衰器３０の入力信号ｓ_ｉｎの振幅値Ｓｉ、受話減衰器３１の出力信号ｒ_ｏｕｔの振幅値Ｒｏの比Ｓｉ／Ｒｏの受話状態における観測値を用いており、遠端側の帰還利得βを推定するために受話減衰器３１の入力信号ｒ_ｉｎの振幅値Ｒｉ、送話減衰器３０の出力信号ｓ_ｏｕｔの振幅値Ｓｏの比Ｒｉ／Ｓｏの送話状態における観測値を用いている。つまり上記公報に記載されているものは、近端側話者の音声信号が存在しない場合、すなわち図６におけるＶｓ＝０のときはＳｉ／Ｒｏ＝αとなり、遠端側話者の音声信号が存在しない場合、すなわち図６におけるＶｒ＝０のときはＲｉ／Ｓｏ＝βとなるという原理に基づいている。
【０００６】
【発明が解決しようとする課題】
上記従来例においては、時刻ｔにおいて求められる受話減衰器３１の入出力信号ｒ_ｉｎ，ｒ_ｏｕｔの振幅値をＲｉ（ｔ），Ｒｏ（ｔ）、送話減衰器３０の入出力信号ｓ_ｉｎ，ｓ_ｏｕｔの振幅値Ｓｉ（ｔ），Ｓｏ（ｔ）とするとＳｉ（ｔ）／Ｒｏ（ｔ）を用いて近端側の帰還利得αを推定し、Ｒｉ（ｔ）／Ｓｏ（ｔ）を用いて遠端側の帰還利得βを推定している。しかし、実際の帰還経路には伝達遅延時間があり、近端側及び遠端側の各帰還経路の伝達遅延時間を各々Ｔα，Ｔβとすると、近端側の帰還利得αはＶｒ＝０におけるＳｉ（ｔ）／Ｒｏ（ｔ−Ｔα）で表され、遠端側の帰還利得βはＶｓ＝０におけるＲｉ（ｔ）／Ｓｏ（ｔ−Ｔβ）で表される。
【０００７】
而して、上記従来例では帰還利得α，βを推定する際に経路の伝達遅延時間Ｔα，Ｔβを考慮していないため、これら伝達遅延時間Ｔα，Ｔβの値が無視できないほど長いときには真の帰還利得と推定値との誤差が大きくなってしまう。従って、そのような場合には信号経路に挿入すべき損失量を精度よく算出することができないと言う問題がある。
【０００８】
本発明は上記問題に鑑みて為されたものであり、その目的とするところは、高い精度で帰還利得を推定し挿入損失量を適切な値に制御することで同時通話性に優れた拡声通話装置を提供することにある。
【０００９】
【課題を解決するための手段】
上記目的を達成するために、請求項１の発明は、集音した音を送話側の音声信号として出力するマイクロホンと、マイクロホンからの音声信号を増幅する第１の増幅手段と、受話側の音声信号に応じて鳴動するスピーカと、スピーカへ出力される音声信号を増幅する第２の増幅手段と、送話側の信号経路に所定量の損失を挿入する送話側損失挿入手段と、受話側の信号経路に所定量の損失を挿入する受話側損失挿入手段と、受話側損失挿入手段及び送話側損失挿入手段の挿入損失量を調整する挿入損失量調整手段とを備えた拡声通話装置において、挿入損失量調整手段は、送話側並びに受話側の音声信号を監視して通話状態を判定する通話状態判定部と、送話側損失挿入手段の入力信号並びに受話側損失挿入手段の出力信号から近端側の帰還利得を推定する近端側帰還利得推定部と、受話側損失挿入手段の入力信号並びに送話側損失挿入手段の出力信号から遠端側の帰還利得を推定する遠端側帰還利得推定部と、近端側帰還利得推定部及び遠端側帰還利得推定部の各推定値に基づいて閉ループに挿入すべき総損失量を算出する総損失量算出部と、通話状態判定部の判定結果と総損失量算出部の算出値に応じて送話側損失挿入手段及び受話側挿入損失手段の各挿入損失量を決定する挿入損失量分配処理部とを具備し、近端側帰還利得推定部は、送話側損失挿入手段の入力信号の時間平均パワーを推定する第１の平均パワー推定部、受話側損失挿入手段の出力信号の時間平均パワーを推定する第２の平均パワー推定部、受話側損失挿入手段の出力点から送話側損失挿入手段の入力点の間の近端側帰還経路にて想定される最大遅延時間における第２の平均パワー推定部の出力の最小値を求める第１の最小値算出部、第１の平均パワー推定部の出力値を第１の最小値算出部の出力値で除算する第１の除算器を有して通話状態判定部により受話状態と判定されたときにのみ各部の処理を更新して成り、遠端側帰還利得推定部は、受話側損失挿入手段の入力信号の時間平均パワーを推定する第３の平均パワー推定部、送話側損失挿入手段の出力信号の時間平均パワーを推定する第４の平均パワー推定部、送話側損失挿入手段の出力点から受話側損失挿入手段の入力点の間の遠端側帰還経路にて想定される最大遅延時間における第４の平均パワー推定部の出力の最小値を求める第２の最小値算出部、第３の平均パワー推定部の出力値を第２の最小値算出部の出力値で除算する第２の除算部を有して通話状態判定部により送話状態と判定されたときにのみ各部の処理を更新して成ることを特徴とし、第１及び第２の最小値算出部にて算出される最小値は帰還経路の伝達遅延時間の影響を低減して高い精度で帰還利得を推定することができ、挿入損失量を適切な値に調整して優れた同時通話性を確保することができる。また、伝達遅延時間の変動に追従して帰還利得を高い精度で推定することが可能である。しかも、帰還利得の推定値が真の値よりも小さくなる場合は少なく、総損失量の不足による信号経路の不安定化を防ぐことができる。
【００１０】
上記目的を達成するために、請求項２の発明は、集音した音を送話側の音声信号として出力するマイクロホンと、マイクロホンからの音声信号を増幅する第１の増幅手段と、受話側の音声信号に応じて鳴動するスピーカと、スピーカへ出力される音声信号を増幅する第２の増幅手段と、送話側の信号経路に所定量の損失を挿入する送話側損失挿入手段と、受話側の信号経路に所定量の損失を挿入する受話側損失挿入手段と、受話側損失挿入手段及び送話側損失挿入手段の挿入損失量を調整する挿入損失量調整手段とを備えた拡声通話装置において、挿入損失量調整手段は、送話側並びに受話側の音声信号を監視して通話状態を判定する通話状態判定部と、送話側損失挿入手段の入力信号並びに受話側損失挿入手段の出力信号から近端側の帰還利得を推定する近端側帰還利得推定部と、受話側損失挿入手段の入力信号並びに送話側損失挿入手段の出力信号から遠端側の帰還利得を推定する遠端側帰還利得推定部と、近端側帰還利得推定部及び遠端側帰還利得推定部の各推定値に基づいて閉ループに挿入すべき総損失量を算出する総損失量算出部と、通話状態判定部の判定結果と総損失量算出部の算出値に応じて送話側損失挿入手段及び受話側挿入損失手段の各挿入損失量を決定する挿入損失量分配処理部とを具備し、近端側帰還利得推定部は、送話側損失挿入手段の入力信号の時間平均パワーを推定する第１の平均パワー推定部、受話側損失挿入手段の出力信号の時間平均パワーを推定する第２の平均パワー推定部、第２の平均パワー推定部の出力を閉ループの最小伝達遅延時間以上の時間だけ遅延させる第１の遅延部、第１の平均パワー推定部の出力値を第１の遅延部の出力値で除算する第１の除算器を有して通話状態判定部により受話状態と判定されたときにのみ各部の処理を更新して成り、遠端側帰還利得推定部は、受話側損失挿入手段の入力信号の時間平均パワーを推定する第３の平均パワー推定部、送話側損失挿入手段の出力信号の時間平均パワーを推定する第４の平均パワー推定部、第４の平均パワー推定部の出力を閉ループの最小伝達遅延時間以上の時間だけ遅延させる第２の遅延部、第３の平均パワー推定部の出力値を第２の遅延部の出力値で除算する第２の除算部を有して通話状態判定部により送話状態と判定されたときにのみ各部の処理を更新して成ることを特徴とし、第１及び第２の遅延部の出力値は帰還経路の伝達遅延時間の影響を低減して高い精度で帰還利得を推定することができ、挿入損失量を適切な値に調整して優れた同時通話性を確保することができる。また、帰還経路の伝達遅延時間が既知であって殆ど変動がないような場合に単純な演算処理により帰還利得を推定することが可能となり、信号処理量の低減化が図れる。
【００１１】
請求項３の発明は、請求項１の発明において、第１及び第２の除算器の出力値のうちから人間の音声の音韻継続時間以上の所定時間内における最小値を算出する第３及び第４の最小値算出部を近端側帰還利得推定部並びに遠端側帰還利得推定部にそれぞれ設け、第３及び第４の最小値算出部の出力値を近端側帰還利得推定部並びに遠端側帰還利得推定部の推定値として成ることを特徴とし、請求項１の発明の作用に加えて、所定時間における帰還利得の推定値の最小値を用いて総損失量を算出するため、所謂ダブルトーク状態において帰還利得の推定値が真の値よりも大きくなるのを抑制することができる。
【００１２】
請求項４の発明は、請求項１の発明において、適応フィルタを有し近端側帰還経路又は遠端側帰還経路の少なくとも何れか一方に設けられるエコーキャンセラと、適応フィルタの係数の収束値からエコーキャンセラを設けた側の帰還経路の利得を近似的に算出する帰還利得近似算出部と、この帰還利得近似算出部の出力値とエコーキャンセラを設けた側の帰還利得推定部の帰還利得推定値との比を求める第３の除算器と、第３の除算器の出力値に応じてエコーキャンセラを設けた側の帰還利得推定部の帰還利得推定値の最新の値と以前の値の何れか一方を選択して総損失量算出部に出力する帰還利得選択部とを備え、この帰還利得選択部は、第３の除算器の出力値が所定の範囲内である場合には最新の帰還利得推定値を選択し、第３の除算器の出力値が所定の範囲を超えた場合には以前の帰還利得推定値を選択して成ることを特徴とし、請求項１の発明の作用に加えて、帰還利得近似算出部の近似的な推定値を参照して適切な推定値を帰還利得選択部で選択しているから、通話路中に衝撃性のノイズが混入した場合等の外乱により生じる推定誤差を低減することができる。
【００１３】
【発明の実施の形態】
（実施形態１）
本実施形態の拡声通話装置は、図１に示すように集音した音を送話側の音声信号として出力するマイクロホン１と、マイクロホン１からの音声信号を増幅する第１の増幅器２と、受話側の音声信号に応じて鳴動するスピーカ３と、スピーカ３へ出力される音声信号を増幅する第２の増幅器４と、送話側の信号経路に所定量の損失を挿入する送話側損失挿入手段たる送話信号減衰器５と、受話側の信号経路に所定量の損失を挿入する受話側損失挿入手段たる受話信号減衰器６と、送話信号減衰器５及び受話信号減衰器６の損失量を調整する挿入損失量調整手段７とを備えている。
【００１４】
また、挿入損失量調整手段７は、図２に示すように送話側並びに受話側の音声信号を監視して通話状態を判定する通話状態判定部８と、送話信号減衰器５の入力信号ｓ_ｉｎ並びに受話信号減衰器６の出力信号ｒ_ｏｕｔから近端側の帰還利得を推定する近端側帰還利得推定部９と、受話信号減衰器６の入力信号ｒ_ｉｎ並びに送話信号減衰器５の出力信号ｓ_ｏｕｔから遠端側の帰還利得を推定する遠端側帰還利得推定部１０と、近端側帰還利得推定部９及び遠端側帰還利得推定部１０の各推定値に基づいて閉ループに挿入すべき総損失量を算出する総損失量算出部１１と、通話状態判定部８の判定結果と総損失量算出部１１の算出値に応じて送話信号減衰器５及び受話信号減衰器６の各挿入損失量を決定する挿入損失量分配処理部１２とを具備している。
【００１５】
さらに近端側帰還利得推定部９は、送話信号減衰器５の入力信号ｓ_ｉｎの時間平均パワーを推定する第１の平均パワー推定部１３と、受話信号減衰器６の出力信号ｒ_ｏｕｔの時間平均パワーを推定する第２の平均パワー推定部１４と、受話信号減衰器６の出力点から送話信号減衰器５の入力点の間の近端側帰還経路にて想定される最大遅延時間における第２の平均パワー推定部１４の出力の最小値を求める第１の最小値算出部１５と、第１の平均パワー推定部１３の出力値を第１の最小値算出部１５の出力値で除算する第１の除算器１６とで構成される。また遠端側帰還利得推定部１０は、受話信号減衰器６の入力信号ｒ_ｉｎの時間平均パワーを推定する第３の平均パワー推定部１７と、送話信号減衰器５の出力信号ｓ_ｏｕｔの時間平均パワーを推定する第４の平均パワー推定部１８と、送話信号減衰器５の出力点から受話信号減衰器６の入力点の間の遠端側帰還経路にて想定される最大遅延時間における第４の平均パワー推定部１８の出力の最小値を求める第２の最小値算出部１９と、第３の平均パワー推定部１７の出力値を第２の最小値算出部１９の出力値で除算する第２の除算部２０とで構成される。なお、挿入損失量調整手段７はＤＳＰ（ＤｉｇｉｔａｌＳｉｇｎａｌＰｒｏｃｅｓｓｏｒ）やＣＰＵ等を用いていて構成することができ、さらには近端側帰還利得推定部９並びに遠端側帰還利得推定部１０のみをアナログ回路で構成してもよい。
【００１６】
通話状態判定部８では送話信号減衰器５の入出力信号ｓ_ｉｎ，ｓ_ｏｕｔ及び受話信号減衰器６の入出力信号ｒ_ｉｎ，ｒ_ｏｕｔを監視し、これらの信号ｓ_ｉｎ，ｓ_ｏｕｔ，ｒ_ｉｎ，ｒ_ｏｕｔのパワーレベルの大小関係並びに音声信号の有無などの情報から通話状態が受話状態、送話状態、遷移状態の何れであるかを判定し、その判定結果を挿入損失量分配処理部１２に与えている。
【００１７】
近端側帰還利得推定部９を構成する第１の平均パワー推定部１３は、整流平滑器や低域通過フィルタ等を用いて送話信号減衰器５への入力信号ｓ_ｉｎの短時間における時間平均パワーＬｓ_ｉｎを求めている。同じく第２の平均パワー推定部１４は、受話信号減衰器６の出力信号ｒ_ｏｕｔの短時間における時間平均パワーＬｒ_ｏｕｔを求めている。また、第１の最小値算出部１５においては、近端側帰還経路において考えられる最大の伝達遅延時間の間、上記時間平均パワーＬｒ_ｏｕｔを観測してその最小値［Ｌｒ_ｏｕｔ］ｍｉｎを求める。第１の除算器１６においては、第１の平均パワー推定部１３の出力Ｌｓ_ｉｎと第１の最小値算出部１５の出力［Ｌｒ_ｏｕｔ］ｍｉｎとの比を算出し、この値を近端側帰還利得の推定値として総損失量算出部１１に出力する。
【００１８】
一方、遠端側帰還利得推定部１０を構成する第３の平均パワー推定部１７は、整流平滑器や低域通過フィルタ等を用いて受話信号減衰器６への入力信号ｒ_ｉｎの短時間における時間平均パワーＬｒ_ｉｎを求めている。同じく第４の平均パワー推定部１８は、送話信号減衰器５の出力信号ｓ_ｏｕｔの短時間における時間平均パワーＬｓ_ｏｕｔを求めている。また、第２の最小値算出部１９においては、遠端側帰還経路において考えられる最大遅延時間の間、上記時間平均パワーＬｓ_ｏｕｔを観測してその最小値［Ｌｓ_ｏｕｔ］ｍｉｎを求める。第２の除算器２０においては、第３の平均パワー推定部１７の出力Ｌｒ_ｉｎと第２の最小値算出部１９の出力［Ｌｓ_ｏｕｔ］ｍｉｎとの比を算出し、この値を遠端側帰還利得の推定値として総損失量算出部１１に出力する。
【００１９】
ここで、近端側帰還利得推定部９における上記推定処理は通話状態判定部８にて通話状態が受話状態と判定されたときに更新され、遠端側帰還利得推定部１０における上記推定処理は通話状態判定部８にて通話状態が送話状態と判定されたときに更新される。
【００２０】
総損失量算出部１１においては、これらの帰還利得推定値から所望の利得余裕（ハウリングマージン）を得るために必要な総挿入損失量を算出し、その値を挿入損失量分配処理部１２に出力する。そして、挿入損失量分配処理部１２では、通話状態判定部８にて判定された通話状態に応じた割合で総挿入損失量を送話信号減衰器５と受話信号減衰器６に分配するように各減衰器５，６の損失量を調整する。
【００２１】
而して、本実施形態では第１及び第２の最小値算出部１５，１９により近端側並びに遠端側の帰還経路において考えられる最大の伝達遅延時間の間、第２及び第４の平均パワー推定部１４，１８の出力値（時間平均パワーＬｒ_ｏｕｔ，Ｌｓ_ｏｕｔ）を観測してその最小値［Ｌｒ_ｏｕｔ］ｍｉｎ，［Ｌｓ_ｏｕｔ］ｍｉｎを求め、その値を用いて帰還利得の推定値を算出しているから、帰還経路の伝達遅延時間の影響を低減して帰還利得の推定精度を向上することができるとともに、上記伝達遅延時間が変動してもそれに追従して高い精度で帰還利得を推定することができる。また、本実施形態では帰還利得の推定値が真の値よりも小さくなる場合は少なく、総損失量の不足による信号経路の不安定化を防ぐことができるという利点もある。
【００２２】
（実施形態２）
本発明の実施形態２における挿入損失量調整手段７のブロック図を図３に示す。本実施形態は、実施形態１の挿入損失量調整手段７における第１及び第２の最小値算出部１５，１９を、それぞれ第１の遅延器２１及び第２の遅延器２２に置き換えたものであり、その他の各部の構成及び動作は実施形態１と共通するので、共通する部分には同一の符号を付して説明を省略する。
【００２３】
第１及び第２の遅延器２１，２２は、それぞれ近端側の帰還経路並びに遠端側の帰還経路の構成要素や諸条件（マイクロホン１−スピーカ３間の距離など）により一意に定まる最短の伝達遅延時間以上の一定時間だけ第２の平均パワー推定部１４及び第４の平均パワー推定部１８の出力信号を遅延させて第１及び第２の除算器１６，２０に出力するものである。すなわち、これら近端側及び遠端側の帰還経路の伝達遅延時間を各々Ｔ_Ｎ，Ｔ_Ｆとすると、時刻ｔにおける近端側の帰還利得推定値はＬｓ_ｉｎ（ｔ）／Ｌｒ_ｏｕｔ（ｔ−Ｔ_Ｎ）、遠端側の帰還利得推定値はＬｒ_ｉｎ（ｔ）／Ｌｓ_ｏｕｔ（ｔ−Ｔ_Ｆ）としてそれぞれ表される。
【００２４】
而して本実施形態によれば、帰還経路の伝達遅延時間が既知であり且つ変動が少ない場合に第１及び第２の遅延器２１，２２で上記伝達遅延時間以上の一定時間だけ第２の平均パワー推定部１４及び第４の平均パワー推定部１８の出力信号を遅延させることにより、伝達遅延時間を考慮して精度よく帰還利得を推定することができる。しかも、第１及び第２の遅延器２１，２２は実施形態１における第１及び第２の最小値算出部１５，１９に比較して構成が簡単であり、コストダウンが図れるという利点がある。
【００２５】
（実施形態３）
本発明の実施形態３における挿入損失量調整手段７のブロック図を図４に示す。本実施形態は、実施形態１における挿入損失量調整手段７に対して、第１及び第２の除算器１６，２０の出力側にそれぞれ第３及び第４の最小値算出部２３，２４を設け、これら第３及び第４の最小値算出部２３，２４の出力を近端側及び遠端側の帰還利得推定値として総損失量算出部１１に入力するものであり、その他の各部の構成及び動作は実施形態１と共通するので、共通する部分には同一の符号を付して説明を省略する。
【００２６】
第３及び第４の最小値算出部２３，２４は、ＲＡＭ等の記憶手段を用いて人間の音声の音韻継続時間以上の一定時間だけ第１及び第２の除算器１６，２０の出力値を保持し、その保持した値の最小値をその時点での帰還利得推定値として抽出し総損失量算出部１１に出力する。このような処理により、送話信号減衰器５の入力信号ｓ_ｉｎ並びに受話信号減衰器６の入力信号ｒ_ｉｎにエコー成分（受話信号減衰器６の出力信号ｒ_ｏｕｔ及び送話信号減衰器５の出力信号ｓ_ｏｕｔが各々近端側及び遠端側の帰還経路を経て回り込む信号成分）以外の成分、すなわち近端側の話者の音声信号、遠端側の話者の音声信号並びに周囲騒音が含まれる場合にそれらの影響が最小となる瞬間に推定された帰還利得を得ることができ、推定精度の向上が図れるのである。
【００２７】
上述のように本実施形態によれば、近端側と遠端側の話者が略同時に話す状態、所謂ダブルトーク状態において帰還利得の推定値が真の値よりも大きくなるのを抑制することができる。
【００２８】
（実施形態４）
本実施形態の拡声通話装置は、図５に示すように送話側の信号経路に所定量の損失を挿入する送話信号減衰器５と、受話側の信号経路に所定量の損失を挿入する受話信号減衰器６と、送話信号減衰器５及び受話信号減衰器６の損失量を調整する挿入損失量調整手段７と、適応フィルタ２５ａを有し近端側帰還経路に設けられるエコーキャンセラ２５と、適応フィルタ２５ａの係数の収束値からエコーキャンセラ２５を設けた近端側の帰還経路の利得を近似的に算出する帰還利得近似算出部２６と、この帰還利得近似算出部２６の出力値と近端側帰還利得推定部９の帰還利得推定値との比を求める第３の除算器２７と、第３の除算器２７の出力値に応じて近端側帰還利得推定部９の帰還利得推定値の最新の値と以前の値の何れか一方を選択して総損失量算出部１１に出力する帰還利得選択部２８とを備えている。なお、図示はしていないが実施形態１と同様にマイクロホン１、第１の増幅器２、スピーカ３並びに第２の増幅器４を備えていることは言うまでもない。また、実施形態１と共通する構成については同一の符号を付して説明を省略する。
【００２９】
エコーキャンセラ２５は、従来周知のように近端側帰還経路のインパルス応答を適応フィルタ２５ａにより同定し、参照信号（受話信号減衰器６の出力信号ｒ_ｏｕｔ）からエコー信号を推定して減算器により相殺して近端側において生じるエコーを消去するものである。適応フィルタ２５ａの係数が収束した状態では、適応フィルタ２５ａの係数は近端側の帰還経路のインパルス応答を近似しているので、その係数値から帰還利得を算出することができる。帰還利得近似算出部２６はエコーキャンセラ２５における適応フィルタ２５ａの係数から近端側帰還利得を算出する。すなわち、適応フィルタ２５ａの係数列をｈ_０，ｈ_１，ｈ_２，…，ｈ_Ｉ−１（Ｉは適応フィルタ２５ａのタップ数）とすると、その２乗和（＝ｈ_０ ^２＋ｈ_１ ^２＋ｈ_２ ^２＋…＋ｈ_Ｉ−１ ^２）から帰還利得を近似的に算出することができる。
【００３０】
第３の除算器２７は、帰還利得近似算出部２６の出力信号と近端側帰還利得推定部９の出力信号との比を求める。帰還利得選択部２８は、第３の除算器２７で求めた上記比の値が所定の範囲内か否かを判断し、上記比の値が所定範囲内であれば近端側帰還利得推定部９の出力信号を選択して総損失量算出部１１に出力し、上記比の値が所定範囲外であればそれ以前の時刻において所定範囲内と判断された上記比の値を選択して総損失量算出部１１に出力する。すなわち、上記比の値が１に近い場合は、近端側帰還利得推定部９の推定値には誤差が少ないと考えられるが、上記比の値が１と離れている場合には誤差が多く含まれている可能性が高いから、上記所定範囲を１を中心とした許容誤差範囲とし、許容誤差範囲は挿入損失量が所望の利得余裕（ハウリングマージン）を得ることができるように決定される。
【００３１】
上述のように本実施形態によれば、エコーキャンセラ２５の適応フィルタ係数から帰還利得近似算出部２６にて算出した帰還利得の推定値を参照して近端側帰還利得推定部９における帰還利得推定値の妥当性を確認するため、通話路中に衝撃性のノイズが混入した場合等の外乱により生じる推定誤差を低減することができる。なお、本実施形態ではエコーキャンセラ２５を近端側の帰還経路のみに設けたが、遠端側の帰還経路のみあるいは近端側と遠端側の両方の帰還経路にエコーキャンセラ２５を設けてもよい。なお、本実施形態が効果を発揮するのは、エコーキャンセラによるエコー消去量が所定のしきい値を越えた場合のみに限る。
【００３２】
【発明の効果】
請求項１の発明は、集音した音を送話側の音声信号として出力するマイクロホンと、マイクロホンからの音声信号を増幅する第１の増幅手段と、受話側の音声信号に応じて鳴動するスピーカと、スピーカへ出力される音声信号を増幅する第２の増幅手段と、送話側の信号経路に所定量の損失を挿入する送話側損失挿入手段と、受話側の信号経路に所定量の損失を挿入する受話側損失挿入手段と、受話側損失挿入手段及び送話側損失挿入手段の挿入損失量を調整する挿入損失量調整手段とを備えた拡声通話装置において、挿入損失量調整手段は、送話側並びに受話側の音声信号を監視して通話状態を判定する通話状態判定部と、送話側損失挿入手段の入力信号並びに受話側損失挿入手段の出力信号から近端側の帰還利得を推定する近端側帰還利得推定部と、受話側損失挿入手段の入力信号並びに送話側損失挿入手段の出力信号から遠端側の帰還利得を推定する遠端側帰還利得推定部と、近端側帰還利得推定部及び遠端側帰還利得推定部の各推定値に基づいて閉ループに挿入すべき総損失量を算出する総損失量算出部と、通話状態判定部の判定結果と総損失量算出部の算出値に応じて送話側損失挿入手段及び受話側挿入損失手段の各挿入損失量を決定する挿入損失量分配処理部とを具備し、近端側帰還利得推定部は、送話側損失挿入手段の入力信号の時間平均パワーを推定する第１の平均パワー推定部、受話側損失挿入手段の出力信号の時間平均パワーを推定する第２の平均パワー推定部、受話側損失挿入手段の出力点から送話側損失挿入手段の入力点の間の近端側帰還経路にて想定される最大遅延時間における第２の平均パワー推定部の出力の最小値を求める第１の最小値算出部、第１の平均パワー推定部の出力値を第１の最小値算出部の出力値で除算する第１の除算器を有して通話状態判定部により受話状態と判定されたときにのみ各部の処理を更新して成り、遠端側帰還利得推定部は、受話側損失挿入手段の入力信号の時間平均パワーを推定する第３の平均パワー推定部、送話側損失挿入手段の出力信号の時間平均パワーを推定する第４の平均パワー推定部、送話側損失挿入手段の出力点から受話側損失挿入手段の入力点の間の遠端側帰還経路にて想定される最大遅延時間における第４の平均パワー推定部の出力の最小値を求める第２の最小値算出部、第３の平均パワー推定部の出力値を第２の最小値算出部の出力値で除算する第２の除算部を有して通話状態判定部により送話状態と判定されたときにのみ各部の処理を更新して成るので、第１及び第２の最小値算出部にて算出される最小値は帰還経路の伝達遅延時間の影響を低減して高い精度で帰還利得を推定することができ、挿入損失量を適切な値に調整して優れた同時通話性を確保することができるという効果がある。また、伝達遅延時間の変動に追従して帰還利得を高い精度で推定することが可能であり、しかも、帰還利得の推定値が真の値よりも小さくなる場合は少なく、総損失量の不足による信号経路の不安定化を防ぐことができるという効果がある。
【００３３】
請求項２の発明は、集音した音を送話側の音声信号として出力するマイクロホンと、マイクロホンからの音声信号を増幅する第１の増幅手段と、受話側の音声信号に応じて鳴動するスピーカと、スピーカへ出力される音声信号を増幅する第２の増幅手段と、送話側の信号経路に所定量の損失を挿入する送話側損失挿入手段と、受話側の信号経路に所定量の損失を挿入する受話側損失挿入手段と、受話側損失挿入手段及び送話側損失挿入手段の挿入損失量を調整する挿入損失量調整手段とを備えた拡声通話装置において、挿入損失量調整手段は、送話側並びに受話側の音声信号を監視して通話状態を判定する通話状態判定部と、送話側損失挿入手段の入力信号並びに受話側損失挿入手段の出力信号から近端側の帰還利得を推定する近端側帰還利得推定部と、受話側損失挿入手段の入力信号並びに送話側損失挿入手段の出力信号から遠端側の帰還利得を推定する遠端側帰還利得推定部と、近端側帰還利得推定部及び遠端側帰還利得推定部の各推定値に基づいて閉ループに挿入すべき総損失量を算出する総損失量算出部と、通話状態判定部の判定結果と総損失量算出部の算出値に応じて送話側損失挿入手段及び受話側挿入損失手段の各挿入損失量を決定する挿入損失量分配処理部とを具備し、近端側帰還利得推定部は、送話側損失挿入手段の入力信号の時間平均パワーを推定する第１の平均パワー推定部、受話側損失挿入手段の出力信号の時間平均パワーを推定する第２の平均パワー推定部、第２の平均パワー推定部の出力を閉ループの最小伝達遅延時間以上の時間だけ遅延させる第１の遅延部、第１の平均パワー推定部の出力値を第１の遅延部の出力値で除算する第１の除算器を有して通話状態判定部により受話状態と判定されたときにのみ各部の処理を更新して成り、遠端側帰還利得推定部は、受話側損失挿入手段の入力信号の時間平均パワーを推定する第３の平均パワー推定部、送話側損失挿入手段の出力信号の時間平均パワーを推定する第４の平均パワー推定部、第４の平均パワー推定部の出力を閉ループの最小伝達遅延時間以上の時間だけ遅延させる第２の遅延部、第３の平均パワー推定部の出力値を第２の遅延部の出力値で除算する第２の除算部を有して通話状態判定部により送話状態と判定されたときにのみ各部の処理を更新して成るので、第１及び第２の遅延部の出力値は帰還経路の伝達遅延時間の影響を低減して高い精度で帰還利得を推定することができ、挿入損失量を適切な値に調整して優れた同時通話性を確保することができるという効果がある。また、帰還経路の伝達遅延時間が既知であって殆ど変動がないような場合に単純な演算処理により帰還利得を推定することが可能となり、信号処理量の低減化が図れるという効果がある。
【００３４】
請求項３の発明は、第１及び第２の除算器の出力値のうちから人間の音声の音韻継続時間以上の所定時間内における最小値を算出する第３及び第４の最小値算出部を近端側帰還利得推定部並びに遠端側帰還利得推定部にそれぞれ設け、第３及び第４の最小値算出部の出力値を近端側帰還利得推定部並びに遠端側帰還利得推定部の推定値として成るので、請求項１の発明の効果に加えて、所定時間における帰還利得の推定値の最小値を用いて総損失量を算出するため、所謂ダブルトーク状態において帰還利得の推定値が真の値よりも大きくなるのを抑制することができるという効果がある。
【００３５】
請求項４の発明は、適応フィルタを有し近端側帰還経路又は遠端側帰還経路の少なくとも何れか一方に設けられるエコーキャンセラと、適応フィルタの係数の収束値からエコーキャンセラを設けた側の帰還経路の利得を近似的に算出する帰還利得近似算出部と、この帰還利得近似算出部の出力値とエコーキャンセラを設けた側の帰還利得推定部の帰還利得推定値との比を求める第３の除算器と、第３の除算器の出力値に応じてエコーキャンセラを設けた側の帰還利得推定部の帰還利得推定値の最新の値と以前の値の何れか一方を選択して総損失量算出部に出力する帰還利得選択部とを備え、この帰還利得選択部は、第３の除算器の出力値が所定の範囲内である場合には最新の帰還利得推定値を選択し、第３の除算器の出力値が所定の範囲を超えた場合には以前の帰還利得推定値を選択して成るので、請求項１の発明の効果に加えて、帰還利得近似算出部の近似的な推定値を参照して適切な推定値を帰還利得選択部で選択しているから、通話路中に衝撃性のノイズが混入した場合等の外乱により生じる推定誤差を低減することができるという効果がある。
【図面の簡単な説明】
【図１】実施形態１の全体構成を示すブロック図である。
【図２】同上における挿入損失量調整手段を示すブロック図である。
【図３】実施形態２における挿入損失量調整手段を示すブロック図である。
【図４】実施形態３における挿入損失量調整手段を示すブロック図である。
【図５】実施形態４の一部省略した全体構成を示すブロック図である。
【図６】従来例を示すブロック図である。
【符号の説明】
１マイクロホン
２第１の増幅器
３スピーカ
４第２の増幅器
５送話信号減衰器
６受話信号減衰器
７挿入損失量調整手段[0001]
TECHNICAL FIELD OF THE INVENTION
The present invention relates to a voice communication device used in homes, buildings, factories, and the like.
[0002]
[Prior art]
2. Description of the Related Art Conventionally, in a loudspeaker apparatus such as an intercom, a telephone, or a PHS, a closed loop is formed due to acoustic feedback from a speaker to a microphone and impedance mismatch in a hybrid circuit (2-4 wire conversion circuit), thereby increasing the gain of the amplifier. If the gain of the closed loop becomes 1 or more due to excessive reason or the like, there is a problem that howling occurs. If the howling occurs, it becomes difficult to continue the call. Therefore, a means for suppressing howling is indispensable. Had become.
[0003]
Therefore, conventionally, it is determined whether the communication state is the reception state or the transmission state by monitoring the transmission and reception signal, and in the reception state, a predetermined loss amount is inserted into the transmission signal, and the transmission state is determined. In the state, a so-called voice switch has been widely used in which a predetermined amount of loss is inserted into a received signal to reduce a closed loop loop transmission gain. In the case of using this voice switch, if the insertion loss is made too large, the quality of the call such as a feeling of disconnection during a call is deteriorated, so that the insertion loss needs to be sufficient to obtain a desired gain margin (howling margin). It is desirable to minimize it.
[0004]
To solve the above problems, feedback gains on the near-end side (terminal side from the position where the voice switch is inserted) and the far-end side (line side than the position where the voice switch is inserted) are estimated as needed. An audio switch that adaptively controls the amount of insertion loss in accordance therewith to minimize the amount of insertion loss while maintaining a constant gain margin is provided (see Japanese Patent Application Laid-Open No. 63-212251). .
[0005]
The principle of the prior art described in the above publication will be described with reference to FIG. When the transmission attenuator 30 is connected to the near-end microphone 1 and the reception attenuator 31 is connected to the speaker 3, the input / output signal r of the reception attenuator 31_in, R_outAre rectified and smoothed to obtain the amplitude values Ri, Ro, and the input / output signal s of the transmission attenuator 30._in, S_outAre rectified and smoothed and are defined as Si and So. The input signal s of the transmission attenuator 30 is used to estimate the near-end feedback gain α._in, The output signal r of the receiving attenuator 31_outOf the received signal in the receiving state of the ratio Si / Ro of the amplitude value Ro of the input signal r of the receiving attenuator 31 to estimate the feedback gain β on the far end side_inAnd the output signal s of the transmission attenuator 30_outThe observed value in the transmission state of the ratio Ri / So of the amplitude value So is used. In other words, in the above-mentioned publication, when the voice signal of the near end speaker does not exist, that is, when Vs = 0 in FIG. 6, Si / Ro = α, and the voice signal of the far end speaker is When there is no such element, that is, when Vr = 0 in FIG. 6, the principle is that Ri / So = β.
[0006]
[Problems to be solved by the invention]
In the above conventional example, the input / output signal r of the reception attenuator 31 obtained at time t_in, R_outAre the amplitude values of Ri (t) and Ro (t) and the input / output signal s of the transmission attenuator 30._in, S_outAssuming that the amplitude values are Si (t) and So (t), the feedback gain α on the near end side is estimated using Si (t) / Ro (t), and the feedback gain α is calculated using Ri (t) / So (t). The end side feedback gain β is estimated. However, the actual feedback path has a propagation delay time. If the transmission delay times of the near-end and far-end feedback paths are Tα and Tβ, respectively, the near-end feedback gain α is Si at Vr = 0. (T) / Ro (t−Tα), and the feedback gain β at the far end is expressed by Ri (t) / So (t−Tβ) at Vs = 0.
[0007]
In the above conventional example, since the propagation delay times Tα and Tβ of the path are not taken into account when estimating the feedback gains α and β, when the values of the propagation delay times Tα and Tβ are too long to be neglected, a true value is obtained. The error between the feedback gain and the estimated value increases. Therefore, in such a case, there is a problem that the loss amount to be inserted into the signal path cannot be calculated with high accuracy.
[0008]
SUMMARY OF THE INVENTION The present invention has been made in view of the above problems, and an object of the present invention is to provide a loudspeaker call excellent in simultaneous communicability by estimating a feedback gain with high accuracy and controlling an insertion loss to an appropriate value. It is to provide a device.
[0009]
[Means for Solving the Problems]
In order to achieve the above object, the invention according to claim 1 includes a microphone that outputs a collected sound as an audio signal on the transmitting side, a first amplifying unit that amplifies an audio signal from the microphone, and a microphone on the receiving side. A speaker that sounds in response to the audio signal, a second amplifying unit that amplifies the audio signal output to the speaker, a transmission-side loss insertion unit that inserts a predetermined amount of loss into a transmission-side signal path, A loudspeaker apparatus comprising: a receiving-side loss inserting unit that inserts a predetermined amount of loss into a signal path on the receiving side; and an insertion-loss amount adjusting unit that adjusts the insertion loss amount of the receiving-side loss inserting unit and the transmitting-side loss inserting unit. In the above, the insertion loss amount adjusting means monitors a voice signal of the transmitting side and the receiving side to determine a talking state, and an input signal of the transmitting side loss inserting means and an output of the receiving side loss inserting means. Near-end return from signal A near-end feedback gain estimating unit for estimating the gain, a far-end feedback gain estimating unit for estimating a far-end feedback gain from an input signal of the receiving-side loss inserting unit and an output signal of the transmitting-side loss inserting unit, A total loss amount calculator that calculates a total loss amount to be inserted into the closed loop based on the respective estimated values of the near-end feedback gain estimator and the far-end feedback gain estimator; An insertion loss distribution processing unit that determines each insertion loss amount of the transmitting-side loss insertion unit and the receiving-side insertion loss unit in accordance with the value calculated by the amount calculation unit. A first average power estimator for estimating the time average power of the input signal of the talker loss insertion means, a second average power estimator for estimating the time average power of the output signal of the receiver loss inserter, and a receiver loss insertion Between the output point of the means and the input point of the transmitting side loss insertion means A first minimum value calculating unit that obtains a minimum value of an output of the second average power estimating unit at a maximum delay time assumed in the near-end side feedback path, and an output value of the first average power estimating unit is set to a first value. It has a first divider that divides by the output value of the minimum value calculating unit, and updates the processing of each unit only when it is determined that the receiving state is determined by the call state determining unit. A third average power estimator for estimating the time average power of the input signal of the receiver-side loss insertion means, a fourth average power estimator for estimating the time-average power of the output signal of the transmitter-side loss insertion means, and transmission. A second value for obtaining the minimum value of the output of the fourth average power estimator at the maximum delay time assumed in the far end feedback path between the output point of the side loss insertion means and the input point of the reception side loss insertion means. The output values of the minimum value calculating unit and the third average power estimating unit A second dividing unit for dividing by an output value of the minimum value calculating unit, and updating the processing of each unit only when it is determined that the transmitting state is determined by the call state determining unit. And the minimum value calculated by the second minimum value calculation unit can reduce the influence of the propagation delay time of the feedback path to estimate the feedback gain with high accuracy, and adjust the insertion loss to an appropriate value. And excellent simultaneous callability can be secured. Further, it is possible to estimate the feedback gain with high accuracy by following the fluctuation of the transmission delay time. In addition, the estimated value of the feedback gain is less likely to be smaller than the true value, and the instability of the signal path due to the shortage of the total loss can be prevented.
[0010]
In order to achieve the above object, the invention according to claim 2 includes a microphone that outputs a collected sound as an audio signal on the transmitting side, a first amplifying unit that amplifies the audio signal from the microphone, and a microphone on the receiving side. A speaker that sounds in response to the audio signal, a second amplifying unit that amplifies the audio signal output to the speaker, a transmission-side loss insertion unit that inserts a predetermined amount of loss into a transmission-side signal path, A loudspeaker apparatus comprising: a receiving-side loss inserting unit that inserts a predetermined amount of loss into a signal path on the receiving side; and an insertion-loss amount adjusting unit that adjusts the insertion loss amount of the receiving-side loss inserting unit and the transmitting-side loss inserting unit. In the above, the insertion loss amount adjusting means monitors a voice signal of the transmitting side and the receiving side to determine a talking state, and an input signal of the transmitting side loss inserting means and an output of the receiving side loss inserting means. Near-end return from signal A near-end feedback gain estimating unit for estimating the gain, a far-end feedback gain estimating unit for estimating a far-end feedback gain from an input signal of the receiving-side loss inserting unit and an output signal of the transmitting-side loss inserting unit, A total loss amount calculator that calculates a total loss amount to be inserted into the closed loop based on the respective estimated values of the near-end feedback gain estimator and the far-end feedback gain estimator; An insertion loss distribution processing unit that determines each insertion loss amount of the transmitting-side loss insertion unit and the receiving-side insertion loss unit in accordance with the value calculated by the amount calculation unit. A first average power estimator for estimating the time average power of the input signal of the talker loss insertion means, a second average power estimator for estimating the time average power of the output signal of the talker loss insertion means, and a second average The output of the power estimator is longer than the minimum propagation delay time of the closed loop. A first delay unit that delays by a time, a first divider that divides an output value of the first average power estimating unit by an output value of the first delay unit, and determines that the communication state is the reception state by the communication state determination unit The far-end feedback gain estimating unit estimates the time-average power of the input signal of the receiving-side loss insertion unit, and the transmitting-side loss A fourth average power estimator for estimating the time average power of the output signal of the insertion means, a second delay unit for delaying the output of the fourth average power estimator by a time longer than the minimum transmission delay time of the closed loop, and a third delay unit Having a second divider for dividing the output value of the average power estimator by the output value of the second delay unit, and updating the processing of each unit only when it is determined that the communication state is the transmission state by the communication state determination unit. Wherein the output values of the first and second delay units are The effect of the propagation delay time on the feedback path can be reduced to estimate the feedback gain with high accuracy, and the amount of insertion loss can be adjusted to an appropriate value to ensure excellent simultaneous communication. Further, when the transmission delay time of the feedback path is known and there is almost no variation, it is possible to estimate the feedback gain by a simple arithmetic processing, and the amount of signal processing can be reduced.
[0011]
According to a third aspect of the present invention, in the first aspect of the present invention, a third and a third value for calculating a minimum value of the output values of the first and second dividers within a predetermined time equal to or longer than the phoneme duration of the human voice. 4 are provided in the near-end feedback gain estimator and the far-end feedback gain estimator, respectively, and the output values of the third and fourth minimum-value calculators are used as the near-end feedback gain estimator and the far-end feedback gain estimator. The feedback gain estimator is characterized in that the total loss is calculated by using the minimum value of the feedback gain estimation value in a predetermined time in addition to the operation of the invention of claim 1. In the talk state, it is possible to suppress the estimated value of the feedback gain from becoming larger than the true value.
[0012]
According to a fourth aspect of the present invention, in the first aspect of the present invention, an echo canceller provided with at least one of a near-end feedback path and a far-end feedback path having an adaptive filter and a convergence value of coefficients of the adaptive filter are provided. A feedback gain approximation calculator for approximately calculating the gain of the feedback path on the side provided with the echo canceller; an output value of the feedback gain approximation calculator and a feedback gain estimation value of the feedback gain estimator on the side provided with the echo canceller; A third divider for calculating the ratio of the feedback divider, and either the latest value or the previous value of the feedback gain estimation value of the feedback gain estimator of the feedback gain estimator provided with the echo canceller in accordance with the output value of the third divider A feedback gain selector for selecting one of the outputs and outputting the same to the total loss calculator, wherein the feedback gain selector selects the latest feedback gain when the output value of the third divider is within a predetermined range. Select an estimate and a third divider If the output value exceeds a predetermined range, the previous feedback gain estimation value is selected, and in addition to the operation of the invention of claim 1, an approximate estimation value of the feedback gain approximation calculation unit is provided. , An appropriate estimation value is selected by the feedback gain selection unit, so that an estimation error caused by disturbance such as a case where impact noise is mixed in a communication path can be reduced.
[0013]
BEST MODE FOR CARRYING OUT THE INVENTION
(Embodiment 1)
As shown in FIG. 1, the loudspeaker apparatus according to the present embodiment includes a microphone 1 that outputs a collected sound as a voice signal on a transmitting side, a first amplifier 2 that amplifies a voice signal from the microphone 1, Speaker 3 that sounds in response to the audio signal on the side, a second amplifier 4 that amplifies the audio signal output to the speaker 3, and loss insertion on the transmission side that inserts a predetermined amount of loss into the signal path on the transmission side. A transmitting signal attenuator 5 as a means, a receiving signal attenuator 6 as a receiving side loss inserting means for inserting a predetermined amount of loss into a signal path on the receiving side, and a loss of the transmitting signal attenuator 5 and the receiving signal attenuator 6. And an insertion loss adjusting means 7 for adjusting the amount.
[0014]
As shown in FIG. 2, the insertion loss adjusting means 7 monitors a voice signal on the transmitting side and a voice signal on the receiving side to determine a communication state, and an input signal of the transmission signal attenuator 5. s_inAnd the output signal r of the received signal attenuator 6._outA near-end feedback gain estimator 9 for estimating a near-end feedback gain from the input signal r of the received signal attenuator 6._inAnd the output signal s of the transmission signal attenuator 5_outAnd a feedback gain estimating unit 10 for estimating the feedback gain on the far end side, and a total loop to be inserted into a closed loop based on the estimated values of the near end feedback gain estimating unit 9 and the far end feedback gain estimating unit 10. The total loss calculation unit 11 for calculating the loss amount, and the insertion loss of the transmission signal attenuator 5 and the reception loss of the reception signal attenuator 6 according to the determination result of the call state determination unit 8 and the value calculated by the total loss calculation unit 11. And an insertion loss amount distribution processing unit 12 for determining the amount.
[0015]
Further, the near-end side feedback gain estimator 9 outputs the input signal s of the transmission signal attenuator 5._inA first average power estimator 13 for estimating the time average power of the received signal, and an output signal r of the received signal attenuator 6_outA second average power estimating unit 14 for estimating the time average power of the received signal attenuator 6, and the maximum delay assumed in the near-end feedback path between the output point of the received signal attenuator 6 and the input point of the transmitted signal attenuator 5. A first minimum value calculating unit 15 for obtaining the minimum value of the output of the second average power estimating unit 14 at time, and an output value of the first average power estimating unit 13 being an output value of the first minimum value calculating unit 15 And a first divider 16 for dividing by. Further, the far-end feedback gain estimating unit 10 receives the input signal r of the received signal attenuator 6._inAnd a third average power estimator 17 for estimating the time average power of the transmission signal attenuator 5_outA fourth average power estimating section 18 for estimating the time average power of the transmission signal attenuator 5 and the maximum delay assumed on the far end side feedback path from the output point of the transmission signal attenuator 5 to the input point of the reception signal attenuator 6. A second minimum value calculating section 19 for obtaining the minimum value of the output of the fourth average power estimating section 18 at time, and an output value of the third average power estimating section 17 as an output value of the second minimum value calculating section 19 And a second division unit 20 for dividing by. The insertion loss adjusting means 7 can be configured using a DSP (Digital Signal Processor), a CPU, or the like. Further, only the near-end feedback gain estimating section 9 and the far-end feedback gain estimating section 10 are analog-only. It may be constituted by a circuit.
[0016]
The input / output signal s of the transmission signal attenuator 5 is determined by the call state determination unit 8._in, S_outAnd the input / output signal r of the reception signal attenuator 6._in, R_outTo monitor these signals_in, S_out, R_in, R_outIt is determined whether the communication state is the receiving state, the transmitting state, or the transition state from the information such as the magnitude relationship of the power levels and the presence or absence of the voice signal, and the determination result is given to the insertion loss amount distribution processing unit 12. I have.
[0017]
The first average power estimating unit 13 forming the near-end side feedback gain estimating unit 9 uses an input signal s to the transmission signal attenuator 5 using a rectifier / smoothing device or a low-pass filter._inTime average power Ls in short time_inSeeking. Similarly, the second average power estimator 14 outputs the output signal r of the received signal attenuator 6._outTime average power Lr in a short time_outSeeking. Further, in the first minimum value calculating section 15, during the maximum transmission delay time conceivable in the near-end side return path, the time average power Lr_outAnd the minimum value [Lr_out] Min. In the first divider 16, the output Ls of the first average power estimating unit 13_inAnd the output [Lr of the first minimum value calculating unit 15_out] Min, and outputs this value to the total loss amount calculation unit 11 as an estimated value of the near-end side feedback gain.
[0018]
On the other hand, the third average power estimating unit 17 constituting the far-end feedback gain estimating unit 10 uses the rectifying smoother, the low-pass filter or the like to input the signal r to the reception signal attenuator 6._inTime average power Lr in a short time_inSeeking. Similarly, the fourth average power estimator 18 outputs the output signal s of the transmission signal attenuator 5._outTime average power Ls in short time_outSeeking. In the second minimum value calculating section 19, the time average power Ls during the maximum delay time conceivable in the far-end feedback path._outAnd the minimum value [Ls_out] Min. In the second divider 20, the output Lr of the third average power estimating unit 17_inAnd the output [Ls of the second minimum value calculation unit 19_out] Min, and outputs this value to the total loss calculating section 11 as an estimated value of the far-end feedback gain.
[0019]
Here, the above estimation processing in the near-end feedback gain estimating unit 9 is updated when the talking state determination unit 8 determines that the talking state is the receiving state. It is updated when the call state determination unit 8 determines that the call state is the transmission state.
[0020]
The total loss calculator 11 calculates the total insertion loss required to obtain a desired gain margin (howling margin) from these feedback gain estimation values, and outputs the calculated value to the insertion loss distribution processor 12. I do. Then, the insertion loss amount distribution processing unit 12 distributes the total insertion loss amount to the transmission signal attenuator 5 and the reception signal attenuator 6 at a rate corresponding to the communication state determined by the communication state determination unit 8. The loss amounts of the attenuators 5 and 6 are adjusted.
[0021]
Thus, in the present embodiment, the first and second minimum value calculators 15 and 19 calculate the second and fourth averages during the maximum transmission delay time conceivable in the near-end and far-end return paths. Output values of the power estimators 14 and 18 (time average power Lr_out, Ls_out) And its minimum value [Lr_out] Min, [Ls_out] Min is calculated and the estimated value of the feedback gain is calculated using the value min. Therefore, the influence of the transmission delay time of the feedback path can be reduced and the estimation accuracy of the feedback gain can be improved. Even if the time fluctuates, the feedback gain can be estimated with high accuracy following the fluctuation. Further, in this embodiment, the estimated value of the feedback gain is less likely to be smaller than the true value, and there is an advantage that instability of the signal path due to shortage of the total loss can be prevented.
[0022]
(Embodiment 2)
FIG. 3 is a block diagram of the insertion loss adjusting means 7 according to the second embodiment of the present invention. In the present embodiment, the first and second minimum value calculators 15 and 19 in the insertion loss adjusting means 7 of the first embodiment are replaced with a first delay unit 21 and a second delay unit 22, respectively. The configuration and operation of each of the other units are common to those of the first embodiment. Therefore, the common parts are denoted by the same reference numerals and description thereof is omitted.
[0023]
The first and second delay units 21 and 22 have the shortest length uniquely determined by components and conditions (distance between the microphone 1 and the speaker 3) of the near-end side return path and the far-end side return path, respectively. The output signals of the second average power estimating unit 14 and the fourth average power estimating unit 18 are delayed by a certain time longer than the transmission delay time and output to the first and second dividers 16 and 20. That is, the transmission delay times of these near-end and far-end feedback paths are set to T_N, T_FThen, the feedback gain estimation value on the near end side at time t is Ls_in(T) / Lr_out(T-T_N), The estimated feedback gain at the far end is Lr_in(T) / Ls_out(T-T_F).
[0024]
Thus, according to the present embodiment, when the transmission delay time of the feedback path is known and the fluctuation is small, the first and second delay units 21 and 22 use the second and third delay units 21 and 22 for a predetermined time longer than the transmission delay time. By delaying the output signals of the average power estimating unit 14 and the fourth average power estimating unit 18, it is possible to accurately estimate the feedback gain in consideration of the transmission delay time. Moreover, the first and second delay units 21 and 22 have a simpler configuration than the first and second minimum value calculators 15 and 19 in the first embodiment, and have an advantage that cost can be reduced.
[0025]
(Embodiment 3)
FIG. 4 is a block diagram of the insertion loss adjusting means 7 according to the third embodiment of the present invention. In the present embodiment, third and fourth minimum value calculation units 23 and 24 are provided on the output sides of the first and second dividers 16 and 20, respectively, with respect to the insertion loss amount adjusting means 7 in the first embodiment. The outputs of the third and fourth minimum value calculators 23 and 24 are input to the total loss calculator 11 as near-end and far-end feedback gain estimation values. Since the operation is common to that of the first embodiment, the common portions are denoted by the same reference numerals and description thereof will be omitted.
[0026]
The third and fourth minimum value calculators 23 and 24 use storage means such as a RAM to calculate the output values of the first and second dividers 16 and 20 for a certain period of time equal to or longer than the phoneme duration of human speech. The minimum value of the held value is extracted as an estimated value of the feedback gain at that time, and is output to the total loss amount calculation unit 11. By such processing, the input signal s of the transmission signal attenuator 5_inAnd the input signal r of the reception signal attenuator 6._inThe echo component (the output signal r of the received signal attenuator 6)_outAnd the output signal s of the transmission signal attenuator 5_outSignal components wrapping around the near-end and far-end return paths, respectively, that is, the sound signal of the near-end speaker, the sound signal of the far-end speaker, and the ambient noise. The feedback gain estimated at the moment when those effects are minimized can be obtained, and the estimation accuracy can be improved.
[0027]
As described above, according to the present embodiment, it is possible to suppress the estimated value of the feedback gain from becoming larger than the true value in a state where the near-end and far-end speakers speak almost simultaneously, that is, a so-called double talk state. Can be.
[0028]
(Embodiment 4)
As shown in FIG. 5, the loudspeaker apparatus according to the present embodiment inserts a predetermined amount of loss into a signal path on a transmitting side and a predetermined amount of loss into a signal path on a receiving side. A receiving signal attenuator 6, an insertion loss adjusting means 7 for adjusting the amounts of loss of the transmitting signal attenuator 5 and the receiving signal attenuator 6, and an echo canceller 25 having an adaptive filter 25a and provided in a near-end side feedback path. A feedback gain approximation calculating unit 26 for approximately calculating the gain of the feedback path on the near end side provided with the echo canceller 25 from the convergence value of the coefficient of the adaptive filter 25a; A third divider 27 for obtaining a ratio of the feedback gain estimation value of the near-end feedback gain estimating unit 9 to the feedback gain estimation value, and a feedback gain estimation of the near-end feedback gain estimating unit 9 according to the output value of the third divider 27 Select either the newest value or the previous value And a feedback gain selection unit 28 for outputting the loss amount calculation unit 11. Although not shown, it goes without saying that the microphone 1, the first amplifier 2, the speaker 3, and the second amplifier 4 are provided as in the first embodiment. The same components as those in the first embodiment are denoted by the same reference numerals, and description thereof is omitted.
[0029]
The echo canceller 25 identifies the impulse response of the near-end feedback path by the adaptive filter 25a as is well known in the art, and outputs the reference signal (the output signal r of the reception signal attenuator 6)._out) To estimate the echo signal and cancel it out by the subtractor to cancel the echo generated on the near end side. When the coefficients of the adaptive filter 25a have converged, the coefficients of the adaptive filter 25a approximate the impulse response of the feedback path on the near end side, so that the feedback gain can be calculated from the coefficient value. The feedback gain approximation calculator 26 calculates the near-end feedback gain from the coefficient of the adaptive filter 25a in the echo canceller 25. That is, the coefficient sequence of the adaptive filter 25a is represented by h₀, H₁, H₂, ..., h_I-1(I is the number of taps of the adaptive filter 25a), and its sum of squares (= h₀ ²+ H₁ ²+ H₂ ²+ ... + h_I-1 ²), The feedback gain can be approximately calculated.
[0030]
The third divider 27 calculates the ratio between the output signal of the feedback gain approximation calculator 26 and the output signal of the near-end feedback gain estimator 9. The feedback gain selector 28 determines whether or not the value of the ratio obtained by the third divider 27 is within a predetermined range. If the value of the ratio is within a predetermined range, the near-end feedback gain estimator 9 is output to the total loss amount calculation unit 11, and if the value of the ratio is out of the predetermined range, the value of the ratio determined to be within the predetermined range at a time before that is selected and the total signal is output. Output to the loss amount calculation unit 11. That is, when the value of the above ratio is close to 1, the estimated value of the near-end feedback gain estimating unit 9 is considered to have a small error, but when the value of the ratio is far from 1, the error is large. Since there is a high possibility of being included, the predetermined range is set as an allowable error range centered at 1, and the allowable error range is determined so that the insertion loss amount can obtain a desired gain margin (howling margin). .
[0031]
As described above, according to the present embodiment, the feedback gain estimation in the near-end feedback gain estimator 9 is performed by referring to the estimated value of the feedback gain calculated by the feedback gain approximation calculator 26 from the adaptive filter coefficient of the echo canceller 25. In order to confirm the validity of the value, it is possible to reduce an estimation error caused by a disturbance such as a case where impact noise is mixed in a communication path. In the present embodiment, the echo canceller 25 is provided only on the return path on the near end side. However, the echo canceller 25 may be provided only on the return path on the far end side or on both the return paths on the near and far end sides. Good. The present embodiment is effective only when the amount of echo cancellation by the echo canceller exceeds a predetermined threshold.
[0032]
【The invention's effect】
According to the first aspect of the present invention, there is provided a microphone that outputs a collected sound as an audio signal on a transmitting side, a first amplifying unit that amplifies an audio signal from the microphone, and a speaker that sounds according to the audio signal on a receiving side. A second amplifying means for amplifying an audio signal output to a speaker; a transmitting-side loss inserting means for inserting a predetermined amount of loss into a transmitting-side signal path; In a loudspeaker apparatus comprising: a receiving-side loss inserting unit for inserting a loss; and an insertion-loss-amount adjusting unit for adjusting an insertion loss amount of the receiving-side loss-inserting unit and the transmitting-side loss-inserting unit, the insertion-loss-amount adjusting unit includes: A call state determination unit for monitoring a voice signal of a transmitting side and a receiving side to determine a communication state; and a feedback gain on a near end side from an input signal of a transmitting side loss inserting unit and an output signal of a receiving side loss inserting unit. To estimate the near-end feedback An estimating unit, a far-end feedback gain estimating unit for estimating a far-end feedback gain from an input signal of the receiving-side loss inserting unit and an output signal of the transmitting-side loss inserting unit; a near-end-side feedback gain estimating unit; A total loss amount calculator for calculating a total loss amount to be inserted into the closed loop based on each estimated value of the end side feedback gain estimator, and a determination result of the call state determination unit and a value calculated by the total loss amount calculator. And an insertion loss amount distribution processing unit that determines each insertion loss amount of the transmission side loss insertion unit and the reception side insertion loss unit, wherein the near-end feedback gain estimating unit includes an input signal of the transmission side loss insertion unit. A first average power estimator for estimating the time average power, a second average power estimator for estimating the time average power of the output signal of the receiver loss insertion means, and a transmission loss from the output point of the reception loss insertion means. Assumed in the near-end return path between the input points of the insertion means A first minimum value calculator for obtaining the minimum value of the output of the second average power estimator at the maximum delay time, and dividing the output value of the first average power estimator by the output value of the first minimum value calculator The processing of each unit is updated only when the call state is determined by the call state determination unit to have the first divider, and the far-end feedback gain estimating unit is configured to input the input signal of the reception-side loss insertion unit. A third average power estimating section for estimating the time average power of the signal, a fourth average power estimating section for estimating the time average power of the output signal of the transmitting side loss inserting means, and receiving from the output point of the transmitting side loss inserting means. A second minimum value calculator for obtaining the minimum value of the output of the fourth average power estimator during the maximum delay time assumed in the far end feedback path between the input points of the side loss insertion means, and a third average The output value of the power estimator is divided by the output value of the second minimum value calculator. Since the processing of each section is updated only when the communication state is determined by the call state determination section to have the second division section for calculating, the first and second minimum value calculation sections calculate the values. The minimum value can reduce the influence of the propagation delay time of the feedback path, estimate the feedback gain with high accuracy, and adjust the insertion loss to an appropriate value to ensure excellent simultaneous communication. This has the effect. In addition, it is possible to estimate the feedback gain with high accuracy by following the fluctuation of the propagation delay time, and the estimated value of the feedback gain is rarely smaller than the true value. There is an effect that instability of the signal path can be prevented.
[0033]
According to a second aspect of the present invention, there is provided a microphone that outputs a collected sound as an audio signal on the transmitting side, a first amplifying unit that amplifies the audio signal from the microphone, and a speaker that sounds according to the audio signal on the receiving side. A second amplifying means for amplifying an audio signal output to a speaker; a transmitting-side loss inserting means for inserting a predetermined amount of loss into a transmitting-side signal path; In a loudspeaker apparatus comprising: a receiving-side loss inserting unit for inserting a loss; and an insertion-loss-amount adjusting unit for adjusting an insertion loss amount of the receiving-side loss-inserting unit and the transmitting-side loss-inserting unit, the insertion-loss-amount adjusting unit includes: A call state determination unit for monitoring a voice signal of a transmitting side and a receiving side to determine a communication state; and a feedback gain on a near end side from an input signal of a transmitting side loss inserting unit and an output signal of a receiving side loss inserting unit. To estimate the near-end feedback An estimating unit, a far-end feedback gain estimating unit for estimating a far-end feedback gain from an input signal of the receiving-side loss inserting unit and an output signal of the transmitting-side loss inserting unit; a near-end-side feedback gain estimating unit; A total loss amount calculator for calculating a total loss amount to be inserted into the closed loop based on each estimated value of the end side feedback gain estimator, and a determination result of the call state determination unit and a value calculated by the total loss amount calculator. And an insertion loss amount distribution processing unit that determines each insertion loss amount of the transmission side loss insertion unit and the reception side insertion loss unit, wherein the near-end feedback gain estimating unit includes an input signal of the transmission side loss insertion unit. A first average power estimator for estimating the time average power, a second average power estimator for estimating the time average power of the output signal of the receiver-side loss insertion means, and an output of the second average power estimator as a closed loop minimum. The first to delay by a time longer than the propagation delay time And a first divider for dividing the output value of the first average power estimating unit by the output value of the first delay unit. A far-end feedback gain estimating unit for estimating a time average power of an input signal of the receiving-side loss insertion unit, a time period of an output signal of the transmitting-side loss insertion unit. A fourth average power estimator for estimating the average power, a second delay unit for delaying the output of the fourth average power estimator by a time longer than the minimum transmission delay time of the closed loop, and an output of the third average power estimator. A second division unit for dividing the value by the output value of the second delay unit, and the processing of each unit is updated only when the communication state is determined to be in the transmission state by the communication state determination unit. The output value of the second delay unit reduces the influence of the propagation delay time of the feedback path. It is possible to estimate the feedback gain with high accuracy and to adjust the insertion loss to an appropriate value to ensure excellent simultaneous communication. Further, when the transmission delay time of the feedback path is known and there is almost no change, it is possible to estimate the feedback gain by simple arithmetic processing, and there is an effect that the amount of signal processing can be reduced.
[0034]
According to a third aspect of the present invention, a third and a fourth minimum value calculating unit for calculating a minimum value within a predetermined time equal to or longer than a phoneme duration of a human voice from output values of the first and second dividers is provided. The output values of the third and fourth minimum value calculation units are provided in the near-end feedback gain estimation unit and the far-end feedback gain estimation unit, respectively, and the output values of the third and fourth minimum value calculation units are estimated by the near-end feedback gain estimation unit and the far-end feedback gain estimation unit. Since the total loss is calculated using the minimum value of the estimated value of the feedback gain at a predetermined time, the estimated value of the feedback gain is true in a so-called double talk state. Has an effect that it can be suppressed from becoming larger than the value of.
[0035]
According to a fourth aspect of the present invention, there is provided an echo canceller which has an adaptive filter and is provided in at least one of the near-end feedback path and the far-end feedback path. A feedback gain approximation calculating unit for approximately calculating the gain of the feedback path; And one of the latest value and the previous value of the feedback gain estimation value of the feedback gain estimating unit provided with the echo canceller in accordance with the output value of the third divider and the total loss. A feedback gain selector for outputting to the quantity calculator, wherein the feedback gain selector selects the latest feedback gain estimation value when the output value of the third divider is within a predetermined range, The output value of the divider of 3 exceeds the predetermined range In this case, since the previous feedback gain estimation value is selected, in addition to the effect of the invention of claim 1, an appropriate estimation value is obtained by referring to the approximate estimation value of the feedback gain approximation calculation unit. Since the selection is made by the selection unit, there is an effect that it is possible to reduce an estimation error caused by disturbance such as a case where impact noise is mixed in a communication path.
[Brief description of the drawings]
FIG. 1 is a block diagram illustrating an overall configuration of a first embodiment.
FIG. 2 is a block diagram showing an insertion loss amount adjusting means in the same as above.
FIG. 3 is a block diagram illustrating an insertion loss adjusting unit according to a second embodiment.
FIG. 4 is a block diagram illustrating an insertion loss amount adjusting unit according to a third embodiment.
FIG. 5 is a block diagram showing an entire configuration of a fourth embodiment in which a part is omitted.
FIG. 6 is a block diagram showing a conventional example.
[Explanation of symbols]
1 microphone
2 First amplifier
3 Speaker
4 Second amplifier
5 Transmission signal attenuator
6 Received signal attenuator
7 Insertion loss adjustment means

Claims

A microphone that outputs the collected sound as an audio signal on the transmitting side, a first amplifying unit that amplifies the audio signal from the microphone, a speaker that sounds according to the audio signal on the receiving side, and output to the speaker Second amplifying means for amplifying an audio signal, transmitting side loss inserting means for inserting a predetermined amount of loss into a transmitting side signal path, and receiving side loss for inserting a predetermined amount of loss into a receiving side signal path In a loudspeaker apparatus comprising an insertion means, and an insertion loss adjusting means for adjusting an insertion loss amount of the receiving-side loss inserting means and the transmitting-side loss inserting means, the insertion loss adjusting means includes a transmitting side and a receiving side. And a near-end feedback unit for estimating a near-end feedback gain from an input signal of a transmitting-side loss inserting unit and an output signal of a receiving-side loss inserting unit. Gain estimator and receiver A far-end feedback gain estimator for estimating a far-end feedback gain from an input signal of the loss insertion means and an output signal of the transmission-side loss insertion means; a near-end feedback gain estimator and a far-end feedback gain estimator A total loss amount calculating unit for calculating a total loss amount to be inserted into the closed loop based on the estimated values of the above, and a transmitting-side loss inserting means according to the determination result of the call state determining unit and the calculated value of the total loss amount calculating unit. And an insertion loss amount distribution processing unit that determines each insertion loss amount of the receiving side insertion loss means, and the near end side feedback gain estimating unit estimates the time average power of the input signal of the transmitting side loss insertion means. A first average power estimating unit, a second average power estimating unit for estimating a time average power of an output signal of the receiving-side loss inserting unit, and an input point of the transmitting-side loss inserting unit from an output point of the receiving-side loss inserting unit. To the maximum delay time assumed in the near-end feedback path between A first minimum value calculating section for obtaining a minimum value of the output of the second average power estimating section, and a first value for dividing the output value of the first average power estimating section by the output value of the first minimum value calculating section. It has a divider and updates the processing of each unit only when it is determined that the receiving state is determined by the call state determining unit.The far-end feedback gain estimating unit determines the time-average power of the input signal of the receiving-side loss insertion unit. A third average power estimating unit for estimating the time-average power of the output signal of the transmitting side loss inserting unit, a receiving side loss inserting unit based on the output point of the transmitting side loss inserting unit. The second minimum value calculator and the third average power estimator, which determine the minimum value of the output of the fourth average power estimator during the maximum delay time assumed in the far end feedback path between the input points A second division unit for dividing the output value by an output value of the second minimum value calculation unit And the processing of each unit is updated only when the call state is determined by the call state determination unit to be the transmission state.

A microphone that outputs the collected sound as an audio signal on the transmitting side, a first amplifying unit that amplifies the audio signal from the microphone, a speaker that sounds according to the audio signal on the receiving side, and output to the speaker Second amplifying means for amplifying an audio signal, transmitting side loss inserting means for inserting a predetermined amount of loss into a transmitting side signal path, and receiving side loss for inserting a predetermined amount of loss into a receiving side signal path In a loudspeaker apparatus comprising an insertion means, and an insertion loss adjusting means for adjusting an insertion loss amount of the receiving-side loss inserting means and the transmitting-side loss inserting means, the insertion loss adjusting means includes a transmitting side and a receiving side. And a near-end feedback unit for estimating a near-end feedback gain from an input signal of a transmitting-side loss inserting unit and an output signal of a receiving-side loss inserting unit. Gain estimator and receiver A far-end feedback gain estimator for estimating a far-end feedback gain from an input signal of the loss insertion means and an output signal of the transmission-side loss insertion means; a near-end feedback gain estimator and a far-end feedback gain estimator A total loss amount calculating unit for calculating a total loss amount to be inserted into the closed loop based on the estimated values of the above, and a transmitting-side loss inserting means according to the determination result of the call state determining unit and the calculated value of the total loss amount calculating unit. And an insertion loss amount distribution processing unit that determines each insertion loss amount of the receiving side insertion loss means, and the near end side feedback gain estimating unit estimates the time average power of the input signal of the transmitting side loss insertion means. A first average power estimator, a second average power estimator for estimating a time average power of an output signal of the receiving-side loss insertion means, and an output of the second average power estimator for a time longer than a minimum transmission delay time of a closed loop. Delay unit delays only the first average A first divider that divides the output value of the power estimation unit by the output value of the first delay unit, and updates the processing of each unit only when the call state is determined to be in the reception state by the call state determination unit; The far-end feedback gain estimating unit estimates a time-average power of the input signal of the receiving-side loss insertion unit, and a fourth average-power estimation unit estimates the time-average power of the output signal of the transmitting-side loss insertion unit. The second delay unit delays the output of the fourth average power estimator and the output of the fourth average power estimator by a time equal to or longer than the minimum transmission delay time of the closed loop. A voice communication device comprising a second divider for dividing by the output value of (i) and updating the processing of each unit only when the communication state determination unit determines that the communication state is the transmission state.

A third and fourth minimum value calculator for calculating a minimum value within a predetermined time equal to or longer than a phoneme duration of a human voice from output values of the first and second dividers is a near-end feedback gain estimator. And the output values of the third and fourth minimum value calculation units are provided as estimation values of the near-end feedback gain estimation unit and the far-end feedback gain estimation unit, respectively. The voice communication device according to claim 1, wherein

An echo canceller provided with at least one of the near-end feedback path and the far-end feedback path having an adaptive filter, and the gain of the feedback path on the side provided with the echo canceller is approximated from the convergence value of the coefficient of the adaptive filter. And a third divider for calculating a ratio between an output value of the feedback gain approximation calculation unit and a feedback gain estimation value of a feedback gain estimation unit provided with the echo canceller. The feedback which selects one of the latest value and the previous value of the feedback gain estimation value of the feedback gain estimation unit on the side provided with the echo canceller according to the output value of the divider, and outputs it to the total loss amount calculation unit A gain selector, wherein the feedback gain selector selects the latest feedback gain estimated value when the output value of the third divider is within a predetermined range, and outputs the output value of the third divider. If the value exceeds the specified range, the Hands-free communication device according to claim 1, wherein the changing made by selecting the gain estimates.