JP3301775B2 - Voice recognition control device - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a speech recognition device, and more particularly to a speech recognition control device for controlling audio / video equipment by speech recognition.

[0002]

2. Description of the Related Art As a means for controlling audio / video equipment (AV equipment) such as a radio and a television, a control device based on voice recognition is used.

FIG. 3 is a schematic block diagram of such a conventional general speech recognition control device. The conventional voice recognition control device includes an AV device (31) that is a controlled unit and a remote controller (30) that is a remote control unit. The remote controller (30) is connected to the AV device (3) via a wireless medium (32).
Send a control signal to 1).

In FIG. 3, reference numeral (301) denotes a microphone to which sound is input, and (302) denotes a microphone (3).
01) is a voice analysis unit that analyzes a voice signal input from step 01) and extracts a time series of feature parameters representing voice characteristics. For example, a spectrum parameter in which sound signal level information is stored by frequency analysis is obtained.

[0005] (303) is a speech section cutout section for cutting out a section (speech section) in which a speech exists in the time series of the characteristic parameter obtained from the speech analysis section (302), and (304) is a speech section cutout section. This is a pattern creation unit that creates an input speech pattern from the characteristic parameter time series, and obtains a speech pattern in which the characteristic pattern is normalized to a specific time series.

Reference numeral (305) denotes a standard pattern memory in which a large number of standard voice patterns are stored in advance as standard voice patterns, and the voice recognition control device shown in FIG. In this case, standard voice patterns obtained by patterning the characteristics of average voices that are common to all speakers are stored for various voices.

[0007] (306) is the voice pattern creation unit (3)
04) is a comparison / determination unit that performs pattern matching between the input voice pattern obtained from step (04) and each of the standard voice patterns in the standard voice pattern memory (305), and detects a standard voice pattern that minimizes the inter-pattern error. And outputs a recognition result signal corresponding to the standard voice pattern.

(307) converts the recognition result signal obtained from the comparison / determination section (306) into a control signal of an AV device (31) such as a television to be controlled and converts the signal into a control signal of the AV device (3).
This is a remote control transmission unit that transmits the data to 1). Transmission from the remote control transmission unit (307) is performed by a wireless medium (32) such as an optical signal such as infrared light, a radio signal, and a magnetic signal.

A control unit (308) receives a control signal transmitted from the remote control transmission unit (307) via the wireless medium (32) and controls the AV equipment main body (310).
9) is a main body receiving unit for transmitting the control signal to 9). The AV device body (310) is an amplifier (311) for generating audio noise such as voice and music from a speaker (312).
Having.

An operation panel (320) is used for operating the remote controller (30) without performing voice recognition. The operation panel includes various buttons and switches necessary for controlling the AV equipment (31). .

FIGS. 4 and 5 are signal diagrams showing a method of cutting out a voice section in a conventional voice recognition control device. FIG. 4 is a signal diagram showing a method of cutting out a signal (V) consisting of only sound in a quiet environment, and FIG. 5 is a signal diagram showing a method of cutting out a signal (S) consisting of sound and music. It is. In these figures, (B) is a reference value for extracting a voice section having a constant value.

The detection of the voice section is usually performed by detecting the start and end of the voice section based on the level value and the fluctuation state of the input voice signal. The simplest method of this type of detection is used. Is a method comprising comparing means for comparing the level of an audio signal with a predetermined threshold, and regarding a time region in which the level of the audio signal exceeds the threshold as a voice section. According to this method, in the example of FIG. 4, the section (tC1 to tC1) where the level (V) of the audio signal exceeds the threshold (B).
2) is detected as a voice section in which voice is generated.

However, in the conventional voice recognition control device as described above, audio noise such as music is always input from the speaker (312) in addition to the voice input from the microphone (301). 5, the level (S) of the audio signal changes high, and the range (tB1 to tB2) detected as an audio section becomes wider than the actual audio section.

As described above, in the conventional technique, when noise is mixed in a voice signal, it is difficult to accurately detect a time domain of voice, and there is a problem that a recognition rate of voice recognition is reduced.

As a speech recognition technique in an environment where such audio noise is present, Japanese Patent Laid-Open No. 3-2 is disclosed.
As described in Japanese Patent No. 33600, a technique for improving the accuracy of voice section detection by changing a reference value for cutting out a voice section in accordance with an output level of an audio noise source is used.

However, when this technique is used in the conventional apparatus as described above, the level information of audio noise generated by the AV equipment must be reflected in the voice recognition unit. Must be integral with the device. However, when the voice recognition unit and the AV device are integrated, the S / N ratio deteriorates due to the influence of audio noise generated by the AV device, which leads to a reduction in the recognition rate. In order to avoid this, a recognition unit can be provided in the remote controller, but in this case, the level information of audio noise from the AV device cannot be reflected on the speech recognition unit.

[0017]

SUMMARY OF THE INVENTION The present invention has been made in view of the above-mentioned conventional inconveniences, and is intended to be applied to a television or the like which generates ambient noise such as voice and music by a remote control unit for voice recognition. It is an object of the present invention to provide a voice recognition control device capable of controlling a control system based on a result of a voice recognition performed by a remote control unit without being affected by audio noise when controlling a control system. And

[0018]

The controlled system of the speech recognition control apparatus according to the present invention includes a level signal transmitting section for transmitting a level signal following a change in the output level of audio noise to a remote control section.

The remote control unit includes a level signal receiving unit that receives the level signal transmitted from the controlled system, and a cutout reference value setting unit that sets a reference value for cutting out a voice section based on the level signal. A voice analysis unit that analyzes an input voice to extract a feature parameter time series of the voice, and detects a voice region using the cut-out reference value, and determines a voice pattern from the feature parameter time series existing in the voice region. And a voice pattern creation unit that creates the sound pattern.

[0020]

According to the speech recognition control apparatus of the present invention, in the controlled system, the level signal transmitting section transmits the level signal following the fluctuation of the output level of the audio noise to the remote control section.

In the remote control section,
A level signal receiving unit receives the level signal transmitted from the controlled system, a cut-out reference value setting unit sets a reference value for cutting out a voice section based on the level signal, and a voice analysis unit analyzes an input voice. And extract a voice feature parameter time series, a voice pattern creation unit detects a voice area using the cutout reference value, and creates a voice pattern based on the feature parameter time series present in the voice area. A comparison / determination unit compares and determines each of the standard patterns in the standard pattern memory with the voice pattern to identify the voice pattern, and a control signal transmitting unit controls a control signal based on a comparison / determination result by the comparison / determination unit. Send to the system.

[0022]

BRIEF DESCRIPTION OF THE DRAWINGS FIG. 1 is a block diagram showing a speech recognition control apparatus according to the present invention.

FIG. 1 is a schematic configuration diagram of a voice recognition control device according to the present invention. The voice recognition control device according to the present invention also includes an AV device (11) as a controlled unit and a remote control (10) as a remote control unit, like the conventional voice recognition control device, and the remote control (10) is wireless. A control signal is sent to the AV device (11) via the medium (12).

On the remote control (10) side of FIG.
Reference numeral 01) denotes a microphone that inputs voice and converts it into an audio signal, and reference numeral 102 denotes a voice analysis unit that analyzes an audio signal input from the microphone 101 and extracts a time series of feature parameters representing the features of the voice. .

(103) is a speech section cutout section for cutting out a section (speech section) in which speech exists in the time series of the characteristic parameter obtained from the speech analysis section (102), and (104) is an AV device ( This is a cut-out reference value setting unit that sets a cut-out reference value for a voice section based on the audio noise level signal sent from 11). The voice section cut-out unit (103) compares the level of the input voice with the cut-out reference value, regards a time region where the level of the input voice exceeds the cut-out reference value as a voice section, and cuts out the voice of this section.

(105) a pattern creating section for creating an input voice pattern from the characteristic parameter time series of the voice section.

Reference numeral (106) denotes a standard pattern memory in which a large number of standard voice patterns are stored in advance as standard voice patterns, and (107) is an input voice pattern obtained from the voice pattern creating section (105) and A comparison / determination unit that performs pattern matching with each standard voice pattern in the standard voice pattern memory (106) to detect the most similar standard voice pattern, and outputs a recognition result signal corresponding to the detected standard voice pattern.

(108) converts the recognition result signal obtained from the comparison / determination unit (107) into a control signal of an AV device (11) such as a television to be controlled and converts the signal into a control signal.
This is a remote control transmission unit that transmits the data to 1).

A remote control receiving unit (109) receives a signal sent from the AV device (11).
Reference numeral (110) denotes a level signal detection unit that detects a level signal of audio noise generated by the AV device from the signal.

Further, on the side of the AV equipment (11) in FIG. 1, (111) is a main body of the AV equipment such as a television or an audio device, (112) is a control unit for controlling the AV equipment main body (111), (113) is a main body receiving unit that receives a control signal transmitted from the remote control transmitting unit (108) and transmits the control signal to the control unit (112).

Reference numeral (114) denotes an amplifier for outputting audio noise generated by the AV device main body (112).
The output from the amplifier (114) is converted into sound by a speaker (11).
5) is output to the external space and sent as a signal to the level signal generator (116). The level signal creation section (116) measures the level of the audio noise signal output from the amplifier (114) to create a level signal.

(117) is a level signal generator (116)
Is a main body transmission unit for transmitting the level signal created in the step (1) to the remote control (10) side. Note that the remote control transmission unit (10
The transmission from 8) and the transmission from the main body transmission section (117) are performed by a wireless medium (12) such as an optical signal such as infrared light, a radio signal, and a magnetic signal.

An operation panel (120) is used for operating the remote controller (10) without performing voice recognition, and includes various buttons and switches necessary for controlling the AV equipment (11). .

FIG. 2 is a signal diagram showing a voice extracting method according to the apparatus of the present invention. In FIG. 2, (S) indicates the level of the audio signal from the microphone (101), which is obtained by adding the level of audio noise to the audio signal (V) in FIG. This is the same as the signal (S) described above. (B) shows a constant voice section cutout reference value, and (A) shows a voice section cutout reference value dynamically changed according to audio noise.

The operation of the voice recognition control device according to the present invention will now be described. Now, the voice recognition control device according to the present embodiment emits voice from the speaker (115) of the AV device (11). Therefore, it is assumed that both the sound for control and the sound emitted from the speaker are input to the microphone (101).

First, on the side of the AV equipment (11), the level signal creating section (116) receives audio noise generated by the AV equipment main body (111) via the amplifier (114) and changes the output level of the audio noise. The tracked level signal is transmitted from the main body transmission section (117) to the remote control (10).

On the remote control (10) side, the AV equipment (1)
The level signal sent from the remote control receiving unit (1)
09) in the level signal detection unit (110). The cut-out reference value setting unit (104) sets a cut-out reference value with reference to the level signal detected here. The cut-out reference value can be considered as a function of the value of the level signal, and can be represented by, for example, an expression such as A = c × (level signal value) + B. Here, c and B are constants. In particular, B is given an optimum value such that stationary noise input from the microphone (101) is not cut out as speech.

Now, when the user enters the microphone (101)
When the input of the voice to the voice is started, the voice analysis unit (10
2) Analyzing an acoustic signal input from the microphone (101) to extract a time series of characteristic parameters representing characteristics of the voice, and obtaining a spectrum parameter storing voice signal level information by frequency analysis.

The voice section clipping section (103) is a section (tA1 to tA2) where the voice signal level (V) from the microphone (101) exceeds the clipping reference value (A) set by the clipping reference value setting section (104). ) Is detected as a voice section. That is, since the audio area is detected by using the cut-out reference value (A) that changes according to the level of audio noise generated by the AV device (11), it is obtained when a constant cut-out reference value (B) is used. Voice section (tB
1 to tB2) rather than the actual voice section (tC1 to tC2).
A voice section close to 2) can be cut out.

Thereafter, a voice pattern creating section (105)
Creates a voice pattern based on the feature parameter time series existing in the voice section among the feature parameter time series obtained from the voice section cutout unit (103). The comparison / determination unit (107) compares and determines each of the standard patterns in the standard pattern memory (106) with the audio pattern to identify the audio pattern, and transmits a control signal based on the comparison / determination result to the remote control transmission unit (108). ) To the AV device (11).

Again on the side of the AV equipment (11), the control unit (1)
12) receives a control signal transmitted from the remote control transmitting section (108) via the main body receiving section (113), and controls the AV equipment main body (111) according to the received control signal.

The remote control (1) is transmitted from the AV device (11).
The transmission of the level signal to (0) is not always performed, but may be performed only when the remote control (10) is operated. When the remote control (10) starts to perform voice recognition, the AV device ( By sending a signal requesting the start of transmission of a level signal to 11) and starting transmission of a level signal from the AV device (11), a speech recognition control device according to the present invention in a more preferred form can be provided.

In the present embodiment, the speech recognition is performed by pattern matching. However, the speech recognition control device of the present invention using a speech recognition method using probability information, fuzzy, or neural network is also possible.

[0044]

As described above, according to the present invention, in a voice recognition control apparatus including a controlled system such as an AV device which generates audio noise and a remote control unit for controlling the same, the controlled system The level information of the audio noise generated is transmitted to the remote control unit, and in the speech recognition in the remote control unit, the reference level for cutting out the voice section using this information is changed according to the input level of the audio noise. Therefore, it is possible to improve the accuracy of cutting out the voice section.

Therefore, when the operator inputs a voice,
Even if audio noise such as voice or music generated by the device is input overlapping with the voice of the operator, it is possible to prevent an extremely lower recognition rate of voice recognition due to the sound generated by the AV device.

[Brief description of the drawings]

FIG. 1 is a schematic configuration diagram of a voice recognition control device according to the present invention.

FIG. 2 is a signal diagram illustrating a voice cutout method according to the apparatus of the present invention.

FIG. 3 is a schematic configuration diagram of a conventional voice recognition control device.

FIG. 4 is a signal diagram showing a clipping method by a conventional voice recognition control device.

FIG. 5 is a signal diagram showing a clipping method by a conventional voice recognition control device.

[Explanation of symbols]

 Reference Signs List 10 remote control 11 AV equipment 12 wireless medium 101 microphone 103 voice section cutout section 104 cutout reference value setting section 112 control section 116 level signal creation section

──────────────────────────────────────────────────続 き Continuation of the front page (51) Int.Cl. ⁷ identification code FI H04Q 9/00 301 311 (58) Field surveyed (Int.Cl. ⁷ , DB name) G10L 11/02 G10L 15/00 G10L 15 / 04 G10L 15/20 G10L 21/02 H04Q 9/00 301 H04Q 9/00 311

Claims (3)

(57) [Claims] 1. A controlled system for generating audio noise,
A remote control unit for controlling the controlled system based on the result of recognizing the input voice. In the voice recognition control device, the controlled system includes a level signal that follows a change in the output level of the audio noise. And a level signal transmitting section for transmitting a level signal transmitted from the controlled system to the remote control section, and a reference for cutting out a voice section based on the level signal. A cut-out reference value setting unit for setting a value, a voice analyzing unit for analyzing an input voice to extract a feature parameter time series of the voice, and detecting a voice region using the cut-out reference value, and with a voice pattern creation section that creates a sound pattern based on the characteristic parameter time series, and the remote control State but to start the voice recognition
At which point the level signal is
Sends a signal requesting the start of transmission, and
A speech recognition control device for starting transmission of a level signal . 2. A voice recognition control device comprising: a remote control unit for issuing a control signal based on a recognition result of an input voice; and a controlled system controlled by the control signal, wherein the remote control unit comprises: A microphone for inputting voice, a voice analyzer for analyzing a sound signal obtained from the microphone to extract a time series of feature parameters of the voice, and a level signal receiver for receiving a level signal transmitted from the controlled system. A clipping reference value setting unit that sets a reference value for clipping audio based on the level signal;
A voice pattern generation unit that detects a voice region using the cut-out reference value and generates a voice pattern based on a time series of characteristic parameters in the voice region; and stores a voice pattern of a plurality of standard voices in advance as a standard pattern. A standard pattern memory, a comparison / determination unit for comparing / determining each standard pattern in the standard pattern memory with the voice pattern and identifying the voice pattern, and a control signal based on a comparison / determination result by the comparison / determination unit. Control signal sending means for sending to a control system, wherein the controlled system is an audio noise generating means for generating audio noise, a control signal receiving unit for receiving a control signal issued from the control signal sending means, A control unit for controlling the controlled system based on the control signal, and an output of audio noise output from the audio noise generating means. A level signal transmitting section for transmitting a level signal which follows the variation in level to the level signal receiver, comprising said remote control unit starts speech recognition status
At which point the level signal is
Sends a signal requesting the start of transmission, and
A speech recognition control device for starting transmission of a level signal . 3. The voice recognition control device according to claim 2, further comprising an operation unit for generating the control signal with respect to the control signal sending unit.