JP3764394B2 - Quality estimation method and apparatus - Google Patents

Description

【００３１】
［スケール変換法］
以上のようにして、客観測定値と主観評価値との対応関係ｍ（ＰＩ）、および同一評価対象に対する各客観測定値のバラツキの大きさｆ（ＭＯＳ）を予め得ておく。ここで対応関係ｍ（ＰＩ）で用いられた主観評価値には、被試験者の心理的距離の違いが含まれているため、本実施の形態では、これを同一評価対象に対する各客観測定値のバラツキの大きさで正規化している。これをスケール変換法という。
【００３２】
そして、客観測定値ＰＩ（２１）と心理的距離を考慮した主観評価値すなわち正規化主観評価値ＣＳＳ（２２）との関係式すなわち正規化対応関係ｇを得ておくことにより、ネットワークから実測された入力客観測定値に対応する所望の主観評価値を推定する際の処理、特に正規化手段１１における正規化処理の効率化を実現している。
【００３３】
以下、スケール変換法について説明する。
このスケール変換法では、数２に示すように、前述した同一評価対象に対する各客観測定値のバラツキの大きさｆ（ＭＯＳ）を重み関数に用いて、対応関係ｍ（ＰＩ）の主観評価値ＭＯＳを、心理的距離が正規化された主観評価値ＣＳＳに変換している。
数２では、「ＭＯＳ＝３」が「普通（まあ良い）」と評価される評点を主観評価値の基準値としているが、主観評価値が「０」から「４」の尺度で表現される場合は、その中央に位置する「２」を基準とすればよい。
【００３４】
【数２】

【００３５】
ここでは、主観評価値でのバラツキの大きさと基準値でのバラツキの大きさの違い、すなわちｆ（ＭＯＳ）とｆ（３）との差を求め、正規化する主観評価値が基準値以上の場合は、当該主観評価値に差を加算し、正規化する主観評価値が基準値を下回る場合は、当該主観評価値から差を減算している。
これにより、基準値を中心に主観評価値の評価尺度が、「ＭＯＳ＝１」および「ＭＯＳ＝５」の両端方向へ延伸され、特にバラツキが少ない両端付近ほど大きく引き延ばされた形となる。
【００３６】
上記数２により変換された主観評価値ＣＳＳと客観測定値ＰＩであるパケットごとの性能情報との関係は、図６のようなグラフとなる。この図６の各プロットから求められた回帰曲線を示す多項式すなわち正規化対応関係ｇ（ＰＩ）が数３である。
なお、図６には、比較として、正規化を行わない重み付けなしの主観評価値の各プロットから求められた回帰曲線も示してある。
【００３７】
【数３】

【００３８】
この正規化対応関係ｇ（ＰＩ）を用いて主観評価値を推定した結果を図７に示す。また、比較のために、正規化を行わずに推定した結果、すなわち重みを付けずに距離のみ変換したものを図８に示す。
図７では、寄与率が０．９３であり、図８では寄与率が０．８９であった。このように、寄与率やグラフからも推定精度が向上しており、特に主観評価値の低い部分での推定精度が向上していることが分かる。
【００３９】
その後、最終的に所望の推定主観評価値：推定ＭＯＳを推定する際には、先の正規化対応関係ｇ（ＰＩ）より算出した値すなわち正規化主観評価値ＣＳＳを、変換手段１２において数４の変換式ｐ（ＣＳＳ）を用いて再変換し、主観評価値を推定する。ここで用いる変数ＭＯＳは、先の数２で用いたＭＯＳと同じ値を用いる。つまり、数２とは逆の加減算を行う。
【００４０】
【数４】

【００４１】
以上の処理を行って、事前に、正規化対応関係ｇ（ＰＩ）を導出しておくことにより、以後では客観測定を行うのみで、入力された客観測定値ＰＩから所望の推定主観評価値：推定ＭＯＳを推定することが可能となる。ただし、その際には、数４の式ｆ（ＭＯＳ）を、「変換された主観評価値ＣＳＳ」が変数となるように、変換式ｈ（ＣＳＳ）を事前に決定しておく必要がある。つまり、数４を数５に変換して用いる。
【００４２】
【数５】

【００４３】
このように、本実施の形態によれば、推定部１０で、入力客観測定値ＰＩに対応する所望の推定主観評価値：推定ＭＯＳを、その入力客観測定値と被試験者の評価尺度上の心理的距離の違いを正規化した推定主観評価値との正規化された推定対応関係に基づき推定するものとし、そのとき、推定対応関係として、予め実測された客観測定値ＰＩと主観評価値ＭＯＳとの対応関係ｍ（ＰＩ）の主観評価値に含まれる心理的距離の違いを、予め実測された同一評価対象に対する各客観測定値のバラツキの大きさｆ（ＭＯＳ）に基づき正規化して得られた対応関係を用いて、心理的距離を考慮した所望の推定主観評価値：推定ＭＯＳを推定するようにした。
これにより、音声系アプリケーションに対する利用者の心理的な距離の違いを正規化して心理的距離を同一尺度にすることにより、知覚上の品質すなわち主観品質をネットワーク品質から推定できる。
【００４４】
また、推定部１０の正規化手段１１で、予め実測により得られた同一評価対象に対する各客観測定値ＰＩのバラツキの大きさ、具体的には客観測定値ＰＩの分散を用いて、主観評価値に含まれる被試験者の心理的距離の違いを正規化した正規化主観評価値ＣＳＳを求めて、客観測定値ＰＩと正規化主観評価値ＣＳＳとの正規化対応関係ｇ（ＰＩ）を前もって用意しておき、この正規化対応関係ｇ（ＰＩ）を用いて入力客観測定値ＰＩに対応する正規化主観評価値ＣＳＳを算出し、変換手段１２で、変換式ｐ（ＣＳＳ）を用いてその正規化主観評価値ＣＳＳを所望の評価尺度における主観評価値へ変換し所望の推定主観評価値：推定ＭＯＳとして出力するようにした。
これにより、予め実測で得られた客観測定値と主観評価値との対応関係で用いた主観評価値に含まれる心理的な距離の違いを、数２で示したような複雑な演算を必要とすることなく正規化できる。
【００４５】
さらに、正規化対応関係ｇ（ＰＩ）は、当該主観評価値に対応するバラツキの大きさｆ（ＭＯＳ）と主観評価値の基準値に対応するバラツキの大きさｆ（３）との違いに基づき、当該主観評価値が基準値以上の場合には当該主観評価値を増加させ、当該主観評価値が基準値を下回る場合は当該主観評価値を減少させることにより得られた正規化主観評価値ＣＳＳを用いるようにした。
これにより、主観評価値のうち評価の高い部分と低い部分での心理的距離が大きく改善されて推定精度が高くなり、特に主観評価値の低い部分で厳密に推定でき、ＶｏＩＰ音声系アプリケーションの利用限界を判断する際に極めて有用である。
【００４６】
［重み付け回帰法］
次に、本発明の第２の実施の形態にかかる品質推定装置について説明する。
前述した第１の実施の形態では、スケール変換法を用いて、入力客観測定値に対応する所望の推定主観評価値：推定ＭＯＳを推定する場合について説明したが、本実施の形態では、客観測定値ＰＩと推定主観評価値：推定ＭＯＳとの対応関係ｗ（ＰＩ）を重み付け回帰により直接導出する重み付け回帰法を用いる場合について説明する。
【００４７】
本実施の形態にかかる品質推定装置の構成は、前述した図１と同様である。ただし、推定部１０の構成は図９となる。図９は、第２の実施の形態にかかる品質推定装置の推定部１０を示す機能ブロック図である。
推定部１０には、回帰推定式ｗ（ＰＩ）に基づき、入力客観測定値ＰＩ（２１）に対応する所望の推定主観評価値：推定ＭＯＳ（２３）を算出する重み付け回帰推定手段１５が設けられている。
【００４８】
回帰推定式ｗ（ＰＩ）を導出するにあたって、前述と同様に、予め主観評価試験を行い、客観測定値と主観評価値との対応関係ｍ（ＰＩ）、および主観評価値に含まれる心理的距離を正規化するための主観評価値の重み関数ｆ（ＭＯＳ）を得ておく。
そして、対応関係ｍ（ＰＩ）を示す各プロットについて重み関数値ｆ（ＭＯＳ）で重み付けを行う。このとき、主観評価値ＭＯＳを持つプロットごとに、その主観評価値ＭＯＳから求められ重み関数値ｆ（ＭＯＳ）を用いて重み付けを行う。このようにして、重み付けを行った各ブロックから回帰曲線を求める。これにより、客観測定値と心理的距離を考慮した所望の推定主観評価値：推定ＭＯＳとの対応関係を示す回帰推定式ｗ（ＰＩ）が得られる。
【００４９】
図１０は、客観評価値ここではパケットごとの性能情報ＰＩとＭＯＳ（変換されていない主観評価値）の関係を示し、その２つをｆ（ＭＯＳ）の重みを付けて多項式により近似した回帰曲線が示されている。その多項式すなわち回帰推定式ｗ（ＰＩ）を数６に示す。
【００５０】
【数６】

【００５１】
このように、事前に回帰推定式ｗ（ＰＩ）を算出することにより、以後、客観測定を行うのみで、客観測定値から主観評価値を推定することが可能となる。この重み付回帰を用いて算出されたｗ（ＰＩ）によって推定主観評価値：推定ＭＯＳを推定した結果を図１１に示す。比較のために、重み付けを行わずに算出した回帰推定式を用いて推定した結果を図１２に示す。
これらより、寄与率からも推定精度が向上していることが分かる。特に、主観値の低い部分での推定精度が向上しており、音声系アプリケーションの品質が悪い箇所を厳密に推定可能であることを意味している。
【００５２】
［品質管理目標値の推定］
次に、本発明の第３の実施の形態にかかる品質推定装置について説明する。
前述した第１および第２の実施の形態では、実測により得られた客観測定値に対応する推定主観評価値を推定する場合について説明したが、本実施の形態では、音声系アプリケーションの品質について、サービス利用限界として指定された指定主観評価値に応じた客観測定値を推定する場合について説明する。
【００５３】
音声系アプリケーションの品質を管理する際、目標とする品質が得られているかどうかをチェックするには、ネットワークから測定された客観測定値で判断するほうが客観的に判断できる。しかし、実際に目標とする品質を決定する場合は、利用者の主観評価値により決定される。
したがって、目標とする主観評価値に対応する客観測定値を求め、その客観測定値により品質を管理すればよい。
【００５４】
このとき、前述したように目標とする主観評価値には、被試験者の評価尺度上の心理的距離の違いが含まれているため、この心理的距離の違いが考慮されていない客観測定値を品質管理に用いた場合、知覚上の品質とのずれが生じる。
本実施の形態では、このような心理的距離の違いを考慮して、目標となる主観評価値に対応する客観測定値を求めるようにしたものである。
【００５５】
本実施の形態にかかる品質推定装置の構成は、前述した図１と同様であるものの、図１３に示すように、推定部１０には、変換手段４２と正規化手段４１とが設けられている。
変換手段４２および正規化手段４１は、前述した図２の変換手段１２および正規化手段１１の逆処理を行う機能手段である。
【００５６】
まず、変換手段１２では、変換式ｐ（ＣＳＳ）を用いて正規化主観評価値ＣＳＳを推定主観評価値：推定ＭＯＳへ変換していたが、変換手段４２では、目標として選択された主観評価値ＭＯＳを正規化主観評価値ＣＳＳへ変換する。
同じく、正規化手段１１では、入力客観測定値ＰＩから正規化主観評価値ＣＳＳを算出していたが、正規化手段４１では、正規化主観評価値ＣＳＳを客観測定値ＰＩへ変換する。
このとき、変換手段４２では、前述した変換式ｐ（ＣＳＳ）の逆変換を行う変換式Ｐ（ＭＯＳ）を用いればよい。また、正規化手段４１では、前述した正規化対応関係ｇ（ＰＩ）の逆変換を行う正規化対応関係Ｇ（ＣＳＳ）を用いればよい。
【００５７】
また、図１３に代えて、図１４に示すような推定部１０を用いてもよい。この推定部１０には、前述した図９の重み付け回帰推定手段１５の逆処理を行う重み付け回帰推定手段４５が設けられている。
重み付け回帰推定手段１５では、入力客観測定値ＰＩから推定主観評価値を推定していたが、重み付け回帰推定手段４５では、主観評価値から客観測定値を推定する。このとき、重み付け回帰推定手段４５では、前述した回帰推定式ｗ（ＰＩ）の逆変換を行う回帰推定式Ｗ（ＭＯＳ）を用いればよい。
【００５８】
本実施の形態では、このようにして、主観評価値に含まれる心理的距離の違いを考慮して、目標となる主観評価値に対応する客観測定値を求めるようにしたので、目標として選択した主観評価値とネットワークの品質管理に用いる客観測定値とにずれが生じなくなり、目標の主観評価値に対応する適切なネットワーク管理指針を示すことが可能となり、主観的な品質を意識した品質管理を実現できる。
【００５９】
なお、以上の各実施の形態では、パケット網で用いられる音声系アプリケーションを例として説明したが、これに限定されるものではない。例えば、動画系アプリケーションなどのリアルタイム系のアプリケーションに対して本発明を広く適用でき、同様の作用効果が得られる。
【００６０】
【発明の効果】
以上説明したように、本発明は、入力客観測定値に対応する所望の推定主観評価値を、その入力客観測定値と推定主観評価値との推定対応関係に基づき推定するものとし、そのとき、推定対応関係として、予め実測された客観測定値と主観評価値との対応関係の主観評価値に含まれる心理的距離の違いを、予め実測された同一評価対象に対する各客観測定値のバラツキの大きさに基づき正規化して得られた対応関係を用いるようにしたので、リアルタイム系アプリケーションに対する利用者の心理的な距離の違いを正規化してその心理的距離を同一尺度にすることができ、知覚上の品質すなわち主観品質をネットワーク品質から推定できる。
【００６１】
また、品質推定装置の推定部で、指定主観評価値に対応する所望の推定客観測定値を、その指定主観評価値と推定客観測定値との推定対応関係に基づき推定するものとし、そのとき、推定対応関係として、予め実測された客観測定値と主観評価値との対応関係の主観評価値に含まれる心理的距離の違いを、予め実測された同一評価対象に対する各客観測定値のバラツキの大きさに基づき正規化して得られた対応関係を用いるようにしたので、リアルタイム系アプリケーションに対する利用者の心理的な距離の違いを正規化してその心理的距離を同一尺度にすることができ、目標とするパケット網管理指標に対する知覚上のずれのない客観測定値を推定できる。
【図面の簡単な説明】
【図１】 本発明の一実施の形態にかかる品質評価装置の構成を示すブロック図である。
【図２】 第１の実施の形態にかかる品質推定装置の推定部を示す機能ブロック図である。
【図３】 主観評価システムの構成例である。
【図４】 無効パケット率とＭＯＳ値との対応関係ｍ（ＰＩ）を示すグラフである。
【図５】 ＭＯＳ値とその９５％信頼区間（分散×係数）の関係ｒ（ＭＯＳ）を示すグラフである。
【図６】 客観測定値ＰＩと正規化主観評価値ＣＳＳとの関係ｇ（ＰＩ）を示すグラフである。
【図７】 ｇ（ＰＩ）で正規化した推定主観評価値と実測主観評価値との関係を示すグラフである。
【図８】 正規化していない推定主観評価値と実測主観評価値との関係を示すグラフである。
【図９】 第２の実施の形態にかかる品質推定装置の推定部を示す機能ブロック図である。
【図１０】 客観測定値ＰＩと推定主観評価値：推定ＭＯＳとの関係ｗ（ＰＩ）を示すグラフである。
【図１１】 ｗ（ＰＩ）で正規化した推定主観評価値と実測主観評価値との関係を示すグラフである。
【図１２】 正規化していない推定主観評価値と実測主観評価値との関係を示すグラフである。
【図１３】 第３の実施の形態にかかる品質推定装置の推定部を示す機能ブロック図である。
【図１４】 第３の実施の形態にかかる品質推定装置の他の推定部を示す機能ブロック図である。
【符号の説明】
１…演算処理部、２…データ入出力部、３…データ蓄積部、４…操作入力部、５…画面表示部、６…内部バス、１０…推定部、１１、４１…正規化手段、１２、４２…変換手段、１５、４５…重み付け回帰推定手段、２１…客観測定値ＰＩ、２２…正規化主観評価値ＣＳＳ、２３…推定主観評価値：推定ＭＯＳ、３１Ａ，３１Ｂ…ＶｏＩＰアプリケーション、３２Ａ，３２Ｂ…電話機、３３Ａ，３３Ｂ…バブ（ＨＵＢ）、３４Ａ，３４Ｂ…パケット監視装置、３５Ａ，３５Ｂ…背景負荷発生装置、３６Ａ，３６Ｂ…ルータ、３７…通信回線。[0001]
BACKGROUND OF THE INVENTION
The present invention relates to a quality estimation method and apparatus, and more particularly to a quality estimation method and apparatus for estimating subjective evaluation values of real-time applications such as VoIP (Voice over IP) in packet networks and objective measurement values for packet network management indices. is there.
[0002]
[Prior art]
For example, quality evaluation for a real-time application such as a VoIP voice application in a packet network is performed by subjective evaluation or objective evaluation. In particular, subjective evaluation is mainstream in order to quantitatively represent the perceptual quality of the user. In recent years, many measuring devices for objective evaluation have been developed, and there are approaches for estimating subjective evaluation values from objective measurement values.
Subjective evaluation is a method in which an application is finally used, and a human being evaluates its quality. For example, for voice applications, it conforms to the opinion evaluation method used for comprehensive quality evaluation of telephone systems. To estimate the subjective evaluation.
[0003]
The opinion evaluation method is a kind of psychometric method called “Method of Successive Categories”. The test subject evaluates the quality of the voice under test in five levels from “very good” to “very bad” in light of the quality of the telephone used in daily life.
Then, a score of 5 to 1 (or 4 to 0) is given to the evaluation, and a MOS value (Mean Opinion Source) is obtained by calculating a weighted average value according to a votership rate by a large number of subjects. (See, for example, Asaya, “Quality design of communication networks”, The Institute of Electronics, Information and Communication Engineers, 1993, Mar, etc.).
[0004]
The subjective evaluation value represented by such a MOS value is a weighted average according to the voting rate of a large number of subjects, and the variance with respect to the evaluation object evaluated as “very good” or “very bad” is small. The evaluation object that is determined to be “ordinary (good)” has a larger variance than the previous evaluation object because the psychological characteristics of the subject are strongly reflected.
Therefore, the subjective evaluation value is regarded as an evaluation scale that holds a difference in psychological distance.
[0005]
[Problems to be solved by the invention]
However, there is no conventional quality evaluation technique that estimates subjective quality from objective measurement values in consideration of such a psychological distance. In addition, there is no technique for estimating subjective quality considering the difference in psychological distance from network quality such as packet quality such as performance information for each packet, or the packet network management index is considered in consideration of such psychological characteristics. None of them estimated objective measurements.
The present invention is to solve such a problem, and can normalize the difference in the psychological distance of the user with respect to the voice system application to make the psychological distance the same measure, It is an object of the present invention to provide a quality estimation method and apparatus capable of estimating a subjective evaluation value from network quality or estimating an objective measurement value for a packet network management index.
[0006]
[Means for Solving the Problems]
In order to achieve such an object, the quality estimation method according to the present invention provides a subjective evaluation of a test subject from objective measurements that can be measured as network performance with respect to the quality of a real-time application used in a packet network. In the quality estimation method used in the quality estimation apparatus for estimating the subjective evaluation value based on the desired estimation subjective evaluation value corresponding to the input objective measurement value, the input objective measurement value and the estimated subjective evaluation in the estimation unit of the quality estimation apparatus The difference between the psychological distances included in the subjective evaluation value of the correspondence relationship between the objective measurement value and the subjective evaluation value measured in advance is the same as the estimated correspondence relationship. The correspondence obtained by normalization based on the magnitude of the variation of each objective measurement value with respect to the evaluation object is used.
[0007]
Specifically, the estimation unit calculates the normalized subjective evaluation value corresponding to the input objective measurement value based on the normalization correspondence between the input objective measurement value and the normalized subjective evaluation value, and the obtained normalization The subjective evaluation value is converted into a subjective evaluation value in a desired evaluation scale and output as an estimated subjective evaluation value. At that time, as a normalized correspondence, each subjective of the correspondence between the objective measurement value measured in advance and the subjective evaluation value For each evaluation value, each normalized subjective evaluation value obtained by normalizing the difference in psychological distance included in the subjective evaluation value using the magnitude of variation corresponding to the subjective evaluation value, and each objective measurement value You may make it use the correspondence relationship.
[0008]
Further, as a normalization correspondence, based on the difference between the magnitude of variation corresponding to the subjective evaluation value and the magnitude of variation corresponding to the reference value of the subjective evaluation value, if the subjective evaluation value is equal to or greater than the reference value, If the subjective evaluation value is increased and the subjective evaluation value falls below the reference value, the normalized subjective evaluation value obtained by decreasing the subjective evaluation value may be used.
[0009]
As another specific example, based on a regression estimation equation obtained by weighting the size of variation corresponding to each subjective evaluation value from each objective measurement value and subjective evaluation value measured in advance by the estimation unit. The estimated subjective evaluation value corresponding to the input objective measurement value may be output.
You may make it use the dispersion | distribution of the said subjective evaluation value as variation.
[0010]
Another quality estimation method according to the present invention is used in a quality estimation apparatus that estimates an objective measurement value according to a designated subjective evaluation value designated as a service usage limit for the quality of a real-time application used in a packet network. In the quality estimation method, the estimation unit of the quality estimation device estimates a desired estimated objective measurement value corresponding to the designated subjective evaluation value based on the estimated correspondence relationship between the designated subjective evaluation value and the estimated objective measurement value, As a relationship, the difference in the psychological distance included in the subjective evaluation value of the correspondence relationship between the objective measurement value and the subjective evaluation value actually measured in advance is represented by the magnitude of variation of each objective measurement value with respect to the same evaluation target actually measured. The correspondence obtained by normalization based on this is used.
[0011]
Moreover, the quality estimation apparatus according to the present invention estimates a subjective evaluation value based on a subjective evaluation of a test subject from objective measurement values that can be measured as network performance for the quality of a real-time application used in a packet network. The quality estimation apparatus includes an estimation unit that estimates a desired estimated subjective evaluation value corresponding to the input objective measurement value based on an estimated correspondence relationship between the input objective measurement value and the estimated subjective evaluation value. As the correspondence relationship, the difference in psychological distance included in the subjective evaluation value of the correspondence relationship between the objective measurement value and the subjective evaluation value measured in advance is the magnitude of variation of each objective measurement value for the same evaluation target that was measured in advance. The correspondence obtained by normalization based on the above is used.
[0012]
As a specific configuration, the estimating unit calculates a normalized subjective evaluation value corresponding to the input objective measurement value based on a normalization correspondence between the input objective measurement value and the normalized subjective evaluation value, and A conversion means for converting the normalized subjective evaluation value obtained by the normalization means into a subjective evaluation value in a desired evaluation scale and outputting it as an estimated subjective evaluation value, and providing an objective measurement that is measured in advance as a normalization correspondence Obtained by normalizing the difference in psychological distance included in the subjective evaluation value for each subjective evaluation value of the correspondence between the value and the subjective evaluation value using the magnitude of the variation corresponding to the subjective evaluation value A correspondence relationship between each normalized subjective evaluation value and each objective measurement value may be used.
[0013]
Further, as a normalization correspondence, based on the difference between the magnitude of variation corresponding to the subjective evaluation value and the magnitude of variation corresponding to the reference value of the subjective evaluation value, if the subjective evaluation value is equal to or greater than the reference value, If the subjective evaluation value is increased and the subjective evaluation value falls below the reference value, the normalized subjective evaluation value obtained by decreasing the subjective evaluation value may be used.
[0014]
As another specific configuration, the estimation unit uses a regression estimation equation obtained by weighting the magnitude of variation corresponding to each subjective evaluation value from objective measurement values and subjective evaluation values actually measured in advance. A weighted regression estimation unit that outputs an estimated subjective evaluation value corresponding to the input objective measurement value may be provided.
You may make it use the dispersion | distribution of the said subjective evaluation value as variation.
[0015]
Another quality estimation apparatus according to the present invention is a quality estimation apparatus that estimates an objective measurement value according to a designated subjective evaluation value designated as a service usage limit for the quality of a real-time application used in a packet network. An estimation unit for estimating a desired estimated objective measurement value corresponding to the evaluation value based on an estimated correspondence between the designated subjective evaluation value and the estimated objective measurement value is provided. Obtained by normalizing the difference in the psychological distance included in the subjective evaluation value of the correspondence relationship between the objective measurement value and the subjective evaluation value based on the size of the variation of each objective measurement value for the same evaluation target actually measured in advance. The corresponding relationship is used.
[0016]
DETAILED DESCRIPTION OF THE INVENTION
Next, embodiments of the present invention will be described with reference to the drawings.
FIG. 1 is a block diagram showing a configuration of a quality estimation apparatus according to an embodiment of the present invention.
This quality estimation apparatus is composed of a computer as a whole, and includes an arithmetic processing unit 1, a data input / output unit 2, a data storage unit 3, an operation input unit 4 and a screen display unit 5, which are connected via an internal bus 6. Are connected to each other.
The arithmetic processing unit 1 includes a microprocessor such as a CPU and its peripheral circuits.
[0017]
The data input / output unit 2 includes data used for the operation of the quality estimation device, for example, objective measurement values actually measured by the measuring device, subjective evaluation values determined by the examinee, and programs executed by the arithmetic processing unit 1. A circuit unit for inputting and outputting.
The data storage unit 3 includes a storage device such as a hard disk or a memory, and various data input from the data input / output unit 2, various data used for the operation of the arithmetic processing unit 1, and a program executed by the arithmetic processing unit 1 Is a circuit unit for storing and storing.
[0018]
The operation input unit 4 is a circuit unit that detects an operation of a keyboard or a mouse and notifies the arithmetic processing unit 1 of the operation.
The screen display unit 5 includes a screen display device such as a CRT or an LCD, and is a circuit unit that displays various information such as a processing result in the arithmetic processing unit 1.
The arithmetic processing unit 1 is provided with an estimation unit 10 that estimates the quality of a voice application as a functional unit. The estimation unit 10 is configured by cooperating hardware resources of the arithmetic processing unit 1 and software resources including programs, and is activated in response to an instruction from the operation input unit 4. Processing is started based on data input from the input / output unit 2 or the operation input unit 4. Then, the processing result is displayed on the screen display unit 5.
[0019]
[First Embodiment]
Next, the quality estimation apparatus according to the first embodiment of the present invention will be described with reference to FIG. FIG. 2 is a functional block diagram showing the estimation unit 10 of the quality estimation apparatus according to the first embodiment.
The quality estimation apparatus of the present embodiment has the configuration of FIG. 1 described above, and the estimation unit 10 is provided with normalization means 11 and conversion means 12 as shown in FIG. .
[0020]
The estimation unit 10 calculates an estimated subjective evaluation value: estimated MOS (23) corresponding to the input objective measurement value PI (21), and the difference between the input objective measurement value PI and the psychological distance on the test subject's evaluation scale. The estimated subjective evaluation value obtained by normalizing: Estimated and estimated based on the normalized estimated correspondence with the estimated MOS.
At this time, as the estimated correspondence relationship, the difference in the psychological distance included in the subjective evaluation value of the correspondence relationship between the objective measurement value and the subjective evaluation value measured in advance is calculated as the value of each objective measurement value for the same evaluation target measured in advance. A desired estimated subjective evaluation value: estimated MOS in consideration of psychological distance is estimated using the correspondence obtained by normalization based on the size of the variation.
[0021]
In the estimation unit 10, such an estimation process is realized using the normalization unit 11 and the conversion unit 12.
The normalizing means 11 inputs the objective measurement based on the normalized correspondence g (PI) indicating the correspondence between the objective measurement value obtained in advance and the normalized subjective evaluation value in which the difference in psychological distance is normalized. This is a functional unit that calculates a normalized subjective evaluation value CSS (Converted Subject Score) (22) corresponding to a value PI (Packet Information) (21).
The conversion unit 12 is a functional unit that converts the normalized subjective evaluation value CSS obtained by the normalization unit 11 into a subjective evaluation value in a desired evaluation scale, and outputs it as a desired estimated subjective evaluation value: estimated MOS (23). is there.
[0022]
The normalized correspondence used by the normalizing means 11 is obtained by using the magnitude of variation corresponding to the subjective evaluation value for each subjective evaluation value for each objective measurement value obtained by the correspondence relationship. It consists of a correspondence between each normalized subjective evaluation value obtained by normalizing the included psychological distance and each original objective measurement value.
[0023]
[Correspondence between objective measurement values and subjective evaluation values]
First, the correspondence relationship m (PI) between the objective measurement value and the subjective evaluation value, obtained in advance by actual measurement, will be described. FIG. 3 shows a configuration example of a measurement system for obtaining a correspondence relationship between objective measurement values and subjective evaluation values.
This correspondence is obtained by conducting a general subjective evaluation test (for example, see Japanese Patent Application No. 2000-247421 by the inventor of the present invention). Hereinafter, an example thereof will be described with reference to FIG.
[0024]
The LAN-A on the packet transmission side includes a VoIP application 31A for executing a real-time application such as VoIP, a telephone 32A connected to the VoIP application 31A, a hub (HUB) 33A constituting the LAN-A, and a VoIP application 31A. A packet monitoring device 34A for monitoring (capturing) a test packet transmitted from the VoIP via the hub 33A, and a router 36A for connecting the WAN communication line 37 and the LAN-A are provided.
[0025]
The LAN-B on the packet receiving side includes a VoIP application 31B for executing a real-time application such as VoIP, a telephone 32B connected to the VoIP application 31B, a hub (HUB) 33B constituting the LAN-B, and a VoIP application 31B. A packet monitoring device 34B for monitoring (capturing) test packets received by the VoIP via the hub 33A and a router 36B for connecting the WAN communication line 37 and the LAN-B are provided.
[0026]
FIG. 4 is a graph showing the correspondence m (PI) between the invalid packet rate and the MOS value. Each regression curve of this graph is obtained by actually testing various network loads by controlling the background load generation devices 35A and 35B of the measurement system shown in FIG. .
As for the MOS value, 20 subjects evaluated each voice file of 224 conditions recorded on the network by receiving the voice from the telephone 32A on the transmitting side and recording it on the telephone 32B on the receiving side. Was used as the MOS value for the audio file.
[0027]
In addition, a packet loss rate indicating the ratio of the number of lost packets to the total number of transmitted packets at this time was calculated, and a regression curve was obtained from these correspondences. In addition to the packet loss rate, a regression curve may be obtained for packet delay fluctuation as an objective measurement value.
[0028]
[Psychological distance of subjective evaluation value]
Next, the psychological distance of the subjective evaluation value will be described.
In the subjective evaluation test, the average of the scores of all the subjects for one evaluation object (for example, one evaluation sound), that is, the subjective evaluation value MOS and variance are obtained for all evaluation objects, as shown in FIG.
In addition, this subjective evaluation test may utilize the test result of the subjective evaluation test for calculating | requiring the correspondence of the objective measurement value mentioned above and a subjective evaluation value, and may implement it separately.
[0029]
According to FIG. 5, the variance of the scores of the evaluation targets whose subjective evaluation values are evaluated near “1” or “5” is smaller than the variance of those evaluated near “3”. Such a difference in dispersion is a difference in psychological distance.
In order to normalize this distance, a regression curve r (MOS) is calculated by performing a polynomial (quadratic) approximation of a 95% confidence interval (variance × coefficient) for the MOS as shown in FIG. The inverse of the regression curve is a subjective evaluation value weighting function f (MOS) for normalizing the distance.
In this way, the magnitude of the variation of each objective measurement value for the same evaluation target, here f (MOS), is obtained in advance. R (MOS) and f (MOS) obtained from FIG.
[0030]
[Expression 1]

[0031]
[Scale conversion method]
As described above, the correspondence relationship m (PI) between the objective measurement value and the subjective evaluation value and the magnitude f (MOS) of variation of each objective measurement value with respect to the same evaluation object are obtained in advance. Here, since the subjective evaluation value used in the correspondence relationship m (PI) includes a difference in the psychological distance of the examinee, this objective measurement value for the same evaluation object is used in the present embodiment. Normalized by the size of the variation. This is called a scale conversion method.
[0032]
Then, by obtaining the relational expression between the objective measurement value PI (21) and the subjective evaluation value considering the psychological distance, that is, the normalized subjective evaluation value CSS (22), that is, the normalized correspondence g, the measurement is actually performed from the network. In addition, the process for estimating a desired subjective evaluation value corresponding to the input objective measurement value, particularly the efficiency of the normalization process in the normalization means 11 is realized.
[0033]
Hereinafter, the scale conversion method will be described.
In this scale conversion method, as shown in Equation 2, the subjective evaluation value MOS of the correspondence relationship m (PI) is calculated by using, as a weighting function, the magnitude f (MOS) of the variation of each objective measurement value for the same evaluation object described above. Is converted into a subjective evaluation value CSS in which the psychological distance is normalized.
In Equation 2, the score at which “MOS = 3” is evaluated as “ordinary (good)” is used as the reference value of the subjective evaluation value, but the subjective evaluation value is expressed on a scale from “0” to “4”. In this case, “2” located at the center may be used as a reference.
[0034]
[Expression 2]

[0035]
Here, the difference between the magnitude of the variation in the subjective evaluation value and the variation in the reference value, that is, the difference between f (MOS) and f (3) is obtained, and the subjective evaluation value to be normalized is greater than or equal to the reference value. In this case, the difference is added to the subjective evaluation value, and when the normalized subjective evaluation value is lower than the reference value, the difference is subtracted from the subjective evaluation value.
As a result, the evaluation scale of the subjective evaluation value centering on the reference value is extended in the both end directions of “MOS = 1” and “MOS = 5”, and is particularly extended near both ends with less variation. .
[0036]
The relationship between the subjective evaluation value CSS converted by Equation 2 and the performance information for each packet, which is the objective measurement value PI, is a graph as shown in FIG. A polynomial representing a regression curve obtained from each plot in FIG. 6, that is, a normalized correspondence g (PI) is Equation 3.
FIG. 6 also shows a regression curve obtained from each plot of unweighted subjective evaluation values without normalization as a comparison.
[0037]
[Equation 3]

[0038]
FIG. 7 shows the result of estimating the subjective evaluation value using this normalized correspondence g (PI). For comparison, FIG. 8 shows the result of estimation without normalization, that is, the result of converting only the distance without weighting.
In FIG. 7, the contribution rate is 0.93, and in FIG. 8, the contribution rate is 0.89. Thus, it can be seen from the contribution rate and the graph that the estimation accuracy is improved, and in particular, the estimation accuracy is improved in the portion where the subjective evaluation value is low.
[0039]
Thereafter, when a desired estimated subjective evaluation value: estimated MOS is finally estimated, a value calculated from the above-described normalized correspondence g (PI), that is, a normalized subjective evaluation value CSS is calculated by the conversion means 12 using Equation 4 Is converted again using the conversion formula p (CSS), and the subjective evaluation value is estimated. The variable MOS used here is the same value as the MOS used in Equation 2 above. In other words, addition / subtraction reverse to Equation 2 is performed.
[0040]
[Expression 4]

[0041]
By performing the above processing and deriving the normalized correspondence g (PI) in advance, it is only necessary to perform objective measurement thereafter, and a desired estimated subjective evaluation value from the inputted objective measurement value PI: The estimated MOS can be estimated. However, in that case, it is necessary to determine the conversion formula h (CSS) in advance so that the expression f (MOS) of Equation 4 is a variable of the “converted subjective evaluation value CSS”. That is, Formula 4 is converted to Formula 5 and used.
[0042]
[Equation 5]

[0043]
As described above, according to the present embodiment, the estimation unit 10 determines the desired estimated subjective evaluation value: estimated MOS corresponding to the input objective measurement value PI, on the input objective measurement value and the test subject's evaluation scale. The difference between psychological distances is estimated based on the normalized estimated correspondence relationship with the normalized estimated subjective evaluation value, and at that time, as the estimated correspondence relationship, the objective measurement value PI and the subjective evaluation value MOS that are actually measured in advance are estimated. Is obtained by normalizing the difference in psychological distance included in the subjective evaluation value of the correspondence relationship m (PI) with the reference value f (MOS) of the variation of each objective measurement value for the same evaluation object actually measured in advance. By using the corresponding relationship, a desired estimated subjective evaluation value: estimated MOS in consideration of psychological distance is estimated.
Accordingly, by normalizing the difference in the psychological distance of the user with respect to the voice system application and making the psychological distance the same scale, the perceptual quality, that is, the subjective quality can be estimated from the network quality.
[0044]
Further, the normalization means 11 of the estimation unit 10 uses the magnitude of the variation of each objective measurement value PI with respect to the same evaluation object obtained by actual measurement in advance, specifically, the subjective evaluation value using the variance of the objective measurement value PI. The normalized subjective evaluation value CSS obtained by normalizing the difference in the psychological distance of the examinee included in the test is obtained, and the normalization correspondence g (PI) between the objective measurement value PI and the normalized subjective evaluation value CSS is prepared in advance. In addition, the normalized subjective evaluation value CSS corresponding to the input objective measurement value PI is calculated using the normalized correspondence relationship g (PI), and the conversion unit 12 uses the conversion formula p (CSS) to calculate the normalized subjective evaluation value CSS. The converted subjective evaluation value CSS is converted into a subjective evaluation value in a desired evaluation scale and output as a desired estimated subjective evaluation value: estimated MOS.
As a result, it is necessary to perform a complicated calculation as shown in Equation 2 for the difference in the psychological distance included in the subjective evaluation value used in the correspondence between the objective measurement value and the subjective evaluation value obtained in advance by actual measurement. Can be normalized without
[0045]
Further, the normalized correspondence g (PI) is based on the difference between the variation size f (MOS) corresponding to the subjective evaluation value and the variation size f (3) corresponding to the reference value of the subjective evaluation value. The normalized subjective evaluation value CSS obtained by increasing the subjective evaluation value when the subjective evaluation value is equal to or higher than the reference value and decreasing the subjective evaluation value when the subjective evaluation value is lower than the reference value. Was used.
As a result, the psychological distance between the high evaluation part and the low evaluation part of the subjective evaluation value is greatly improved, and the estimation accuracy is improved. This is extremely useful in determining the limits.
[0046]
[Weighted regression method]
Next, a quality estimation apparatus according to the second embodiment of the present invention will be described.
In the first embodiment described above, a case has been described in which a desired estimated subjective evaluation value: estimated MOS corresponding to an input objective measurement value is estimated using the scale conversion method, but in this embodiment, objective measurement is performed. A case where a weighted regression method for directly deriving a correspondence w (PI) between a value PI and an estimated subjective evaluation value: estimated MOS by weighted regression will be described.
[0047]
The configuration of the quality estimation apparatus according to the present embodiment is the same as that shown in FIG. However, the structure of the estimation part 10 becomes FIG. FIG. 9 is a functional block diagram illustrating the estimation unit 10 of the quality estimation apparatus according to the second embodiment.
The estimation unit 10 is provided with weighted regression estimation means 15 for calculating a desired estimated subjective evaluation value: estimated MOS (23) corresponding to the input objective measurement value PI (21) based on the regression estimation formula w (PI). ing.
[0048]
In deriving the regression estimation formula w (PI), a subjective evaluation test is performed in advance in the same manner as described above, the correspondence m (PI) between the objective measurement value and the subjective evaluation value, and the psychological distance included in the subjective evaluation value. The weight function f (MOS) of the subjective evaluation value for normalizing is obtained in advance.
Then, each plot showing the correspondence m (PI) is weighted with the weight function value f (MOS). At this time, for each plot having the subjective evaluation value MOS, weighting is performed using the weight function value f (MOS) obtained from the subjective evaluation value MOS. In this way, a regression curve is obtained from each weighted block. As a result, a regression estimation formula w (PI) indicating a correspondence relationship between an objective measured value and a desired estimated subjective evaluation value: estimated MOS in consideration of a psychological distance is obtained.
[0049]
FIG. 10 shows the relationship between objective evaluation values, here, performance information PI for each packet and MOS (subjective subjective evaluation value), and a regression curve obtained by approximating the two with a weight of f (MOS) and a polynomial expression. It is shown. The polynomial, that is, the regression estimation formula w (PI) is shown in Equation 6.
[0050]
[Formula 6]

[0051]
As described above, by calculating the regression estimation formula w (PI) in advance, it is possible to estimate the subjective evaluation value from the objective measurement value only by performing objective measurement thereafter. FIG. 11 shows a result of estimating the estimated subjective evaluation value: estimated MOS by w (PI) calculated using this weighted regression. For comparison, FIG. 12 shows the result of estimation using a regression estimation formula calculated without weighting.
From these, it can be seen that the estimation accuracy is also improved from the contribution rate. In particular, the estimation accuracy in the portion with a low subjective value is improved, which means that it is possible to accurately estimate a portion where the quality of the voice application is poor.
[0052]
[Estimation of quality control target value]
Next, a quality estimation apparatus according to the third embodiment of the present invention will be described.
In the first and second embodiments described above, a case has been described in which an estimated subjective evaluation value corresponding to an objective measurement value obtained by actual measurement is estimated. However, in the present embodiment, the quality of a voice application is described. The case where the objective measurement value according to the designated subjective evaluation value designated as the service use limit is estimated will be described.
[0053]
When managing the quality of a voice system application, it can be objectively judged by judging from objective measured values measured from the network in order to check whether or not the target quality is obtained. However, when the target quality is actually determined, it is determined by the subjective evaluation value of the user.
Therefore, an objective measurement value corresponding to the target subjective evaluation value may be obtained, and quality may be managed based on the objective measurement value.
[0054]
At this time, as described above, the target subjective evaluation value includes the difference in the psychological distance on the test subject's evaluation scale, so the objective measurement value in which the difference in the psychological distance is not taken into consideration. When used for quality control, a deviation from perceptual quality occurs.
In the present embodiment, an objective measurement value corresponding to a target subjective evaluation value is obtained in consideration of such a difference in psychological distance.
[0055]
Although the configuration of the quality estimation apparatus according to the present embodiment is the same as that of FIG. 1 described above, as shown in FIG. 13, the estimation unit 10 is provided with conversion means 42 and normalization means 41. .
The converting means 42 and the normalizing means 41 are functional means for performing the reverse processing of the converting means 12 and the normalizing means 11 shown in FIG.
[0056]
First, the converting means 12 converts the normalized subjective evaluation value CSS into the estimated subjective evaluation value: estimated MOS using the conversion formula p (CSS), but the converting means 42 selects the subjective evaluation value selected as the target. The MOS is converted into a normalized subjective evaluation value CSS.
Similarly, the normalization means 11 calculates the normalized subjective evaluation value CSS from the input objective measurement value PI, but the normalization means 41 converts the normalized subjective evaluation value CSS into the objective measurement value PI.
At this time, the conversion means 42 may use a conversion formula P (MOS) that performs the inverse conversion of the conversion formula p (CSS) described above. Further, the normalization means 41 may use a normalization correspondence G (CSS) that performs the inverse transformation of the normalization correspondence g (PI) described above.
[0057]
Moreover, it may replace with FIG. 13 and may use the estimation part 10 as shown in FIG. The estimation unit 10 is provided with a weighted regression estimation unit 45 that performs a reverse process of the weighted regression estimation unit 15 of FIG. 9 described above.
The weighted regression estimation means 15 estimates the estimated subjective evaluation value from the input objective measurement value PI, but the weighted regression estimation means 45 estimates the objective measurement value from the subjective evaluation value. At this time, the weighted regression estimation means 45 may use a regression estimation formula W (MOS) that performs inverse transformation of the above-described regression estimation formula w (PI).
[0058]
In this embodiment, in this way, the objective measurement value corresponding to the target subjective evaluation value is obtained in consideration of the difference in the psychological distance included in the subjective evaluation value. There is no gap between the subjective evaluation value and the objective measurement value used for network quality control, and it is possible to provide an appropriate network management guideline corresponding to the target subjective evaluation value. realizable.
[0059]
In each of the above embodiments, a voice application used in a packet network has been described as an example. However, the present invention is not limited to this. For example, the present invention can be widely applied to real-time applications such as moving image applications, and similar effects can be obtained.
[0060]
【The invention's effect】
As described above, the present invention estimates the desired estimated subjective evaluation value corresponding to the input objective measurement value based on the estimated correspondence between the input objective measurement value and the estimated subjective evaluation value, As the estimated correspondence, the difference in the psychological distance included in the subjective evaluation value of the correspondence between the objective measurement value measured in advance and the subjective evaluation value is represented by a large variation in each objective measurement value for the same evaluation target measured in advance. Since the correspondence relation obtained by normalization based on the above is used, the difference in the user's psychological distance to the real-time application can be normalized to make the psychological distance the same scale. Quality, that is, subjective quality can be estimated from network quality.
[0061]
In addition, the estimation unit of the quality estimation device estimates a desired estimated objective measurement value corresponding to the designated subjective evaluation value based on the estimated correspondence relationship between the designated subjective evaluation value and the estimated objective measurement value, As the estimated correspondence, the difference in the psychological distance included in the subjective evaluation value of the correspondence between the objective measurement value measured in advance and the subjective evaluation value is represented by a large variation in each objective measurement value for the same evaluation target measured in advance. Since the correspondence relationship obtained by normalization based on this is used, the difference in the user's psychological distance to the real-time application can be normalized to make the psychological distance the same scale, It is possible to estimate an objective measurement value with no perceptual deviation from the packet network management index.
[Brief description of the drawings]
FIG. 1 is a block diagram showing a configuration of a quality evaluation apparatus according to an embodiment of the present invention.
FIG. 2 is a functional block diagram showing an estimation unit of the quality estimation apparatus according to the first embodiment.
FIG. 3 is a configuration example of a subjective evaluation system.
FIG. 4 is a graph showing a correspondence relationship m (PI) between an invalid packet rate and a MOS value.
FIG. 5 is a graph showing a relationship r (MOS) between a MOS value and its 95% confidence interval (dispersion × coefficient).
FIG. 6 is a graph showing a relationship g (PI) between an objective measurement value PI and a normalized subjective evaluation value CSS.
FIG. 7 is a graph showing a relationship between an estimated subjective evaluation value normalized by g (PI) and an actual measured subjective evaluation value.
FIG. 8 is a graph showing a relationship between an estimated subjective evaluation value that is not normalized and an actual subjective evaluation value;
FIG. 9 is a functional block diagram illustrating an estimation unit of a quality estimation apparatus according to a second embodiment.
FIG. 10 is a graph showing a relationship w (PI) between an objective measurement value PI and an estimated subjective evaluation value: estimated MOS;
FIG. 11 is a graph showing a relationship between an estimated subjective evaluation value normalized by w (PI) and an actual measured subjective evaluation value.
FIG. 12 is a graph showing a relationship between an estimated subjective evaluation value that is not normalized and an actual subjective evaluation value;
FIG. 13 is a functional block diagram illustrating an estimation unit of a quality estimation apparatus according to a third embodiment.
FIG. 14 is a functional block diagram showing another estimation unit of the quality estimation apparatus according to the third embodiment.
[Explanation of symbols]
DESCRIPTION OF SYMBOLS 1 ... Operation processing part, 2 ... Data input / output part, 3 ... Data storage part, 4 ... Operation input part, 5 ... Screen display part, 6 ... Internal bus, 10 ... Estimation part, 11, 41 ... Normalization means, 12 , 42 ... conversion means, 15 and 45 ... weighted regression estimation means, 21 ... objective measurement value PI, 22 ... normalized subjective evaluation value CSS, 23 ... estimated subjective evaluation value: estimation MOS, 31A, 31B ... VoIP application, 32A, 32B ... Telephone, 33A, 33B ... HUB, 34A, 34B ... Packet monitoring device, 35A, 35B ... Background load generating device, 36A, 36B ... Router, 37 ... Communication line.

Claims (12)

In the quality estimation method used in the quality estimation device that estimates the subjective evaluation value based on the subjective evaluation of the test subject from the objective measurement value that can be measured as the network performance, regarding the quality of the real-time system application used in the packet network,
The estimation unit of the quality estimation device estimates a desired estimated subjective evaluation value corresponding to the input objective measurement value based on the estimated correspondence relationship between the input objective measurement value and the estimated subjective evaluation value,
As the estimated correspondence, the difference in the psychological distance included in the subjective evaluation value of the correspondence between the objective measurement value measured in advance and the subjective evaluation value is represented by the variation of each objective measurement value for the same evaluation target measured in advance. A quality estimation method using a correspondence obtained by normalization based on a size. The quality estimation method according to claim 1,
In the estimation unit,
A normalized subjective evaluation value corresponding to the input objective measurement value is calculated based on a normalization correspondence between the input objective measurement value and the normalized subjective evaluation value,
The obtained normalized subjective evaluation value is converted into a subjective evaluation value in a desired evaluation scale and output as the estimated subjective evaluation value,
For each subjective evaluation value of the corresponding relationship between the objective measurement value and the subjective evaluation value measured in advance, the normalized correspondence relationship is set to the subjective evaluation value using the magnitude of the variation corresponding to the subjective evaluation value. A quality estimation method comprising a correspondence relationship between each normalized subjective evaluation value obtained by normalizing a difference in the included psychological distance and each objective measurement value. The quality estimation method according to claim 2,
The normalized correspondence is based on the difference between the variation corresponding to the subjective evaluation value and the variation corresponding to the reference value of the subjective evaluation value, when the subjective evaluation value is greater than or equal to the reference value. Uses the normalized subjective evaluation value obtained by increasing the subjective evaluation value and, if the subjective evaluation value is lower than the reference value, decreasing the subjective evaluation value. The quality estimation method according to claim 1,
Based on the regression estimation formula obtained by weighting the size of the variation corresponding to each subjective evaluation value from each objective measurement value and subjective evaluation value actually measured in advance by the estimation unit, the input objective A quality estimation method comprising outputting an estimated subjective evaluation value corresponding to a measured value. In the quality estimation method of Claims 1-4,
A quality estimation method using variance of the subjective evaluation value as the variation. In the quality estimation method used in the quality estimation device that estimates the objective measurement value according to the designated subjective evaluation value specified as the service usage limit for the quality of the real-time system application used in the packet network,
The estimation unit of the quality estimation device estimates a desired estimated objective measurement value corresponding to the designated subjective evaluation value based on the estimated correspondence relationship between the designated subjective evaluation value and the estimated objective measurement value,
As the estimated correspondence, the difference in the psychological distance included in the subjective evaluation value of the correspondence between the objective measurement value measured in advance and the subjective evaluation value is represented by the variation of each objective measurement value for the same evaluation target measured in advance. A quality estimation method using a correspondence obtained by normalization based on a size. In the quality estimation device that estimates the subjective evaluation value based on the subjective evaluation of the test subject from the objective measurement value that can be measured as the network performance for the quality of the real-time system application used in the packet network,
An estimation unit that estimates a desired estimated subjective evaluation value corresponding to the input objective measurement value based on an estimated correspondence relationship between the input objective measurement value and the estimated subjective evaluation value;
The estimation unit calculates, as the estimated correspondence relationship, a difference in psychological distance included in the subjective evaluation value of the correspondence relationship between the objective measurement value measured in advance and the subjective evaluation value, for each objective for the same evaluation target measured in advance. A quality estimation apparatus using a correspondence relationship obtained by normalizing based on a variation in measured values. The quality estimation apparatus according to claim 7, wherein
The estimation unit includes
Normalization means for calculating a normalized subjective evaluation value corresponding to the input objective measurement value based on a normalization correspondence between the input objective measurement value and the normalized subjective evaluation value;
Conversion means for converting the normalized subjective evaluation value obtained by the normalization means into a subjective evaluation value in a desired evaluation scale and outputting the estimated subjective evaluation value,
For each subjective evaluation value of the corresponding relationship between the objective measurement value and the subjective evaluation value measured in advance, the normalized correspondence relationship is obtained by using the magnitude of the variation corresponding to the subjective evaluation value. A quality estimation apparatus comprising a correspondence relationship between each normalized subjective evaluation value obtained by normalizing a difference in the included psychological distance and each objective measurement value. The quality estimation apparatus according to claim 8, wherein
The normalized correspondence is based on the difference between the variation corresponding to the subjective evaluation value and the variation corresponding to the reference value of the subjective evaluation value, when the subjective evaluation value is greater than or equal to the reference value. Increases the subjective evaluation value, and when the subjective evaluation value is lower than the reference value, uses a normalized subjective evaluation value obtained by decreasing the subjective evaluation value. The quality estimation apparatus according to claim 7, wherein
The estimation unit is configured to measure the input objective measurement based on a regression estimation equation obtained by weighting the size of the variation corresponding to each subjective evaluation value from objective measurement values and subjective evaluation values actually measured in advance. A quality estimation apparatus comprising weighted regression estimation means for outputting an estimated subjective evaluation value corresponding to a value. In the quality estimation apparatus of Claims 7-10,
A quality estimation apparatus using variance of the subjective evaluation value as the variation. In the quality estimation device that estimates the objective measurement value according to the specified subjective evaluation value specified as the service usage limit for the quality of the real-time system application used in the packet network,
An estimation unit that estimates a desired estimated objective measurement value corresponding to the designated subjective evaluation value based on an estimated correspondence relationship between the designated subjective evaluation value and the estimated objective measurement value;
The estimation unit calculates, as the estimated correspondence relationship, a difference in psychological distance included in the subjective evaluation value of the correspondence relationship between the objective measurement value measured in advance and the subjective evaluation value, for each objective for the same evaluation target measured in advance. A quality estimation apparatus using a correspondence relationship obtained by normalizing based on a variation in measured values.

Images