JPH0486700A - Voice recognizer - Google Patents

Abstract

PURPOSE:To recognize input voices with good noise resistance and high performance by forming the component corresponding to the signal sound from a signal sound generator intruded as an ambient noise and subtracting this signal component from the input signal from a microphone. CONSTITUTION:This device has an external sound correcting section 14 which is corrected and inputted with the output signal of a signal sound generator 1 and controls the correction to the signal sound of the signal sound generator 1 in the external sound correcting section 14 according to the result of the comparison between the signal sound corrected and inputted via this external sound correcting section 14 and the signal inputted via the microphone 2. The difference between the signal sound (the corrected output signal from the signal sound generator 1) corrected and inputted in the external sound correcting section 14 and the signal inputted via the microphone 2 is determined and this differential signal is recognized as the input voice. The input voice is recognized with the high performance in this way without being affected by the ambient noise. The noise resistance is thus enhanced.

Description

[Detailed Description of the Invention] [Objective of the Invention] (Field of Industrial Application) The present invention provides high-performance recognition of audio signals input in an environment where a signal sound from a signal sound generating device exists as ambient noise. The present invention relates to a speech recognition device that can perform speech recognition.

(Prior Art) Recently, there has been an increasing demand for voice recognition devices as highly natural man-machine interfaces. Particularly in recent years, with the development of information processing technology and semiconductor manufacturing technology, voice recognition devices have been made smaller and cheaper, and the range of their use and application (fields) is expanding more and more. However, in terms of practical application, there are still some issues that need to be resolved, such as noise resistance.

In other words, a pigeon house that uses a voice recognition device in a noisy environment,
Detecting and recognizing the chest noise as voice would be a fatal mistake. Furthermore, the voice is buried in ambient noise, resulting in problems such as the inability to accurately recognize the input voice. Therefore, in the past, when fishing on a boat, efforts were made to use a voice recognition device only in relatively quiet places (environments), or to input voice using a close-talk type microphone that does not easily pick up surrounding noise. ing.

However, when listening to the playback sound from an audio device (signal sound generating device) and controlling its operation by voice input, it is undeniable that the signal sound from the audio device acts as ambient noise to the input sound. , it is essentially impossible to realize a so-called quiet environment. Therefore, in such a case, it is necessary to input voice using the above-mentioned close-talk type microphone. However, this type of microphone is mostly used by being attached directly to the user's (speaker's) body, making it cumbersome to handle. Moreover, the audio input environment changes significantly depending on the level of the output sound from the audio equipment (signal playback level) that acts as noise, so even if a close-talking microphone is used, mixing of surrounding noise can be completely prevented. There are some problems that are difficult. In other words, since the signals from the audio equipment to be controlled directly act as ambient noise, there are problems such as the fact that even if various control information is input by voice, it is very difficult to recognize it.

(Problem to be solved by the invention) In this way, in the past, even if an attempt was made to control the operation of an audio device using voice recognition, for example, the signal sound output from the audio device would directly act as noise on the input voice. Therefore, it is difficult to improve noise resistance and recognize input speech with high performance, and there are problems such as the need to use a close-talk type microphone that is not easy to use. Moreover, even if a close-talk type microphone is used, there are problems such as that it is difficult to stably ensure sufficient noise resistance depending on the level of the output sound from the audio equipment.

The present invention has been made in consideration of these circumstances, and its purpose is to provide a highly noise-resistant system for input audio in an environment where output sounds from signal sound generating equipment such as audio equipment are present. The object of the present invention is to provide a highly practical speech recognition device capable of high performance recognition.

[Structure of the Invention] (Means for Solving the Problems) The present invention provides a method for detecting audio signals input through a microphone in an environment where a signal sound output from a signal sound generating device such as an audio device exists as ambient noise. The speech recognition device includes an external sound correction section that corrects and inputs the output signal of the signal sound generating device, and the signal sound corrected and inputted through the external sound correction section and input through the microphone. The external sound correction section controls the correction of the signal sound in the signal sound generating device according to the comparison result with the received signal, and in response to such control, the external sound correction section inputs the correction input. A difference between the signal sound (corrected output signal from the signal sound generating device) and the signal input through the microphone is determined, and this difference signal is recognized as the input sound. It is.

(Function) According to the present invention, by correcting and inputting the output signal of the signal sound generating device, a component corresponding to the signal sound from the signal sound generating device that is mixed as ambient noise from the microphone is created. Since this signal component is subtracted from the input signal from the microphone, it is possible to effectively cancel the signal sound output from the signal sound generating device that is mixed into the audio signal as ambient noise.

As a result, it is possible to prevent malfunctions and erroneous recognition of voice recognition processing caused by the signal tone output from the signal tone generating device, and to sufficiently improve the recognition performance. This makes it possible to perform voice recognition effectively and with high performance in an environment where ambient noise exists as ambient noise.

(Embodiment) Hereinafter, a speech recognition device according to an embodiment of the present invention will be described with reference to the drawings.

FIG. 1 is a schematic configuration diagram of an embodiment of the apparatus, in which ``■'' is an audio device that outputs a signal sound through a speaker 2. In FIG. The audio device 1 controls the output of a signal tone (for example, the reproduction output of a musical tone such as voice or music) under the control of a device control section 3, and functions as one of the signal tone generating devices. still,
The signal tone generating device is not limited to the audio device 1 described above, but may also be, for example, a public address device or the like.

The voice recognition device according to the present invention can detect signals from a signal sound generating device such as an audio device or a public address system to a speaker 2.
In an environment where a signal sound is output through a microphone, the voice uttered by a speaker is input through a microphone, and this is attempted to be recognized with high performance.The system is configured as follows.

That is, the microphone 11 is for inputting the voice uttered by the speaker, and the sound signal inputted through the microphone 11 is appropriately processed by a preamplifier or the like, and then processed as described below. are supplied to a comparator 12 and a subtracter 13, respectively.

In this way, the sound inputted from the microphone 11 is mixed with the signal sound from the audio device 1 outputted from the speaker 2 or ambient noise when fishing on a boat.

That is, if the signal sound outputted from the audio device 1 through the speaker 2 is A, the signal sound A goes around to the microphone 11 as environmental noise N, and is transmitted to the microphone 11 along with the voice Sk uttered by the speaker. Input via . Such an input signal (Sk +Nk)
is supplied to the comparator 12 and the subtracter 13, respectively.

On the other hand, the external sound correction section 14 directly inputs the signal sound Ak outputted from the audio device 1 or the speaker 2 from the audio device 1, and controls the signal level (amplitude) under the control of the correction control section 15. ) and phase. The audio equipment 1 corrected by the external sound correction unit 14 under correction control by the correction control unit 15
The output signal ak from the comparator 12 and the subtractor 1
3, respectively.

Therefore, the comparator 12 receives the corrected signal sound a and the input signals (S, +N,
), and, for example, calculates the difference component and provides it to the correction control section 15.

The correction control unit 15 uses such a difference component (Sk + Nkah)
Accordingly, the signal sound A is adjusted by the external sound correction unit 14 so that, for example, this difference component (Si +Nhah) is minimized.
, the amount of correction for , is controlled.

By the feedback system represented by the correction loop of the external sound correction section 14 from the comparator 12 via the correction control section 15, the signal sound A in the external sound correction section 14,
For example, a correction signal a approximately equal to the ambient noise N is determined.

Therefore, the subtracter 13 calculates the difference between the signal sound a corrected and inputted as described above and the input signal (Si +N,) inputted via the microphone 11 as (Si + Nh ak). This is then given to the speech recognition section 16 for speech recognition processing. By generating a signal for speech recognition processing in this way, the signal becomes (Sk+Nk-ah)-+Sk, and the speech recognition unit 16 recognizes the speech S, in which the ambient noise N, is canceled by the correction signal ak. It will be processed. The voice S-1 that has been recognized by the voice recognition unit 16
According to the recognition result, for example, the operation of the device control section 3 is controlled, and the output of the signal sound Ak from the audio device 1 is controlled.

That is, according to this embodiment, for example, as schematically shown in FIG. 2, when the audio Sk is uttered and input from the microphone 11 as shown in FIG. 2(a), the audio device 1 When the signal sound A, as shown by the solid line in Figure 2(b), is output into the vocalization environment of the voice Sk, a part of it is output as external noise Nk, as shown by the dashed line in Figure 2(b). The sound goes around to the microphone 11 and is input together with the sound Sk.

As a result, the signal input from the microphone 11 becomes a signal (Si + Nk) in which the voice Sk and the external noise Nk are superimposed, as shown in FIG. 2(c).

In such a human signal environment, the external sound correction section 1
4 inputs the signal S output from the audio equipment 1, corrects its amplitude and phase, and generates and outputs a correction signal ak as shown in FIG. 2(d). Then, by controlling the correction amount by the feedback loop system described above, the signal (Si
+Nh) and the correction signal ak, and the comparison output (difference) is minimized. Minimizing this comparison output means removing external noise N, which is an unwanted component in the signal (St +N,) input via the microphone 11, and This shows that the correction is made to be approximately equal to the external noise N.

The correction signal a corrected in this way is given to the subtracter 13, and the signal (S
By subtracting the external noise N, as shown in FIG. 2(e), the sound S,
It becomes possible to extract only the components of and provide them to the speech recognition unit 16.

As a result, the signal sound Ak output from the audio device 1
Even when the voice Sk is uttered and the voice is recognized and processed in an environment where the signal tone Ak exists, the external noise N, which wraps around the signal tone Ak and is superimposed on the voice S, can be effectively canceled. It becomes possible to recognize the voice sk effectively and with high performance. Furthermore, even if the voice Sk is not input as a voice, the external noise Nk will not be mistakenly recognized, making it possible to prevent such erroneous recognition operations.

Note that here, the comparator 12 and the subtracter 13 are provided independently,
Either the correction control for external sounds and the removal of noise components from the signal input from the microphone 11 are performed separately, or the comparator 12 and the subtractor 13 are integrated, and the noise components from the input signal are removed while the correction control for external sounds is controlled. It may also be removed.

Of course, it is also possible to perform these processes entirely digitally. Furthermore, the control over the correction of external sound can be performed only when the conditions (state) of the signal output from the audio device 1 change, and in other cases, the amount of correction can be kept constant. . Alternatively, a test mode for determining the amount of correction may be set, and after determining the amount of correction by outputting a signal tone that makes it easy to control the amount of correction from the audio device 1, the mode may be returned to the normal voice recognition mode. . In addition, the output signal of the voice recognition unit 16 and the device control unit 3
It is also possible to construct a control system so as to determine the correction amount in the external sound correction section I4 using the output signal from the external sound correction section I4. In addition, the present invention can be implemented with various modifications without departing from the gist thereof.

[Effects of the Invention] As explained above, according to the present invention, when a voice is uttered in an environment where a signal tone is output from a predetermined signal tone generating device, and the voice is recognized and processed, Even if the signal sound acts as noise in the surrounding environment, the input speech can be recognized with high performance without being affected by the surrounding noise, and its noise resistance can be made sufficiently high. can. Further, unlike the conventional method, there is no need to prepare a close-talk type microphone, and therefore, great practical effects such as ease of use can be achieved.

[Brief explanation of the drawing]

FIG. 1 is a schematic configuration diagram of a speech recognition device according to an embodiment of the present invention, and FIG. 2 is a diagram showing signals of various parts for explaining the operation of the embodiment device. ■... Audio equipment, 3... Equipment control unit, 11.
...Microphone, 12...Comparator, 13...Subtractor, 14...External sound correction section, 15...Correction control section, 16...Speech recognition section. Applicant's agent Patent attorney Takehiko Suzue

Claims (1)

[Scope of Claims] A speech recognition device that recognizes a voice signal input through a microphone in an environment where a signal sound output from a signal sound generating device exists as ambient noise, comprising: an output signal of the signal sound generating device. an external sound correction unit that corrects and inputs the signal, a comparator that compares the signal sound corrected and input through the external sound correction unit with the signal input via the microphone, and an output of the comparator that a correction control unit that controls correction of the signal sound in the signal sound generating device in the external sound correction unit, and a signal sound corrected and input in the external sound correction unit and a signal input via the microphone; What is claimed is: 1. A speech recognition device comprising: a subtracter that obtains a difference between the subtractor and a speech recognition unit that recognizes the difference signal obtained through the subtractor as input speech.