TWI770867B - voice recognition device - Google Patents

Abstract

The present invention discloses a speech recognition device, which comprises at least one position acquisition device, a directional sound pickup device, a noise suppressor and a speech recognition processor. The position acquisition device is sequentially coupled to the directional radio device, the noise suppressor and the speech recognition processor. The position acquisition device obtains the physical voice position of the sound source, and outputs the voice position to the directional sound pickup device, so that the directional sound pickup device receives the voice signal generated by the sound source according to the voice position. The noise suppressor eliminates the noise of the voice signal according to the noise model corresponding to the voice position, so as to generate a voice recognition signal. The speech recognition processor receives the speech recognition signal, and generates an operation signal accordingly, thereby improving the accuracy of the speech recognition.

Description

voice recognition device

The present invention relates to a recognition device, and more particularly, to a speech recognition device.

As the voice recognition function becomes more and more mature, various multimedia devices will use the voice recognition device as an input device, such as smart assistants in mobile phones, voice control devices in vehicles, and smart home appliances, adding new colors to technological life in a new way. Users can interact with the device directly without the need for buttons or contacts.

At present, most of the voice recognition systems equipped with voice recognition are applied to personal devices. The device can use directional microphones or limit the sound collection range and situation to achieve better sound collection and recognition effects. If it is far away, such as in the car, it is easy to be affected by noise or feedback, and if it is in public equipment, it will also have the problem of operation interference. For example, when the first operator needs to continuously interact with the public equipment, if the second operator intentionally or unintentionally generates a voice signal to compete for the operation right, the first operator's human-machine interaction experience will be poor. In addition, in an environment where noise is more likely to be generated, and when it is not possible to limit the sound collection range or move the user's position, it will result in poor speech recognition rate and difficulty in system operation. Some of the voice recognition functions commonly found in vehicles currently use the Android Auto system. By saying OK Google, or pressing and holding the voice command button on the steering wheel, it starts to receive voice commands for operation. In general car voice applications are mostly aimed at driving needs: such as making calls, navigating, controlling music playback, or controlling the thermostat system. Functionally, these needs are all one-way demands and will not last for a long time. operation, etc. Most of the central control systems of commercial vehicles directly use non-directional microphones, so the sound reception effect is easily affected by the feedback of the speakers and noise interference. If the directional microphones on the market are used, it will be difficult for passengers in other positions except driving. operate. Today's existing conference radio products use 360-degree omnidirectional radio to meet the needs of conference radio, and most of them are high-sensitivity microphones. The purpose is to accurately receive the voice of all participants in the conference room. These devices focus on noise filtering to keep the sound clear. After receiving the sound, most of them will use dynamic noise reduction (Digital Noise Reduction, DNR), sound gain control, or other related methods to increase the intensity of the human voice and the ability to pick up the sound. However, due to the need to receive the voices of all conference participants as much as possible, there is no need for directivity, so it will not pick up the sound in the direction of each operator, and will not particularly suppress other voices.

Therefore, the present invention proposes a speech recognition device in order to solve the above-mentioned problems.

The present invention provides a voice recognition device, which reduces the frequency of grabbing the control right and improves the operability of the public equipment when the voice signal controls the public equipment, and in a complex and airtight environment, improves the quality of sound, the directionality of sound and the reduction of noise. Noise function to improve the accuracy of speech recognition.

In one embodiment of the present invention, a speech recognition device is provided, which includes at least one position acquisition device, a directional sound pickup device, a noise suppressor and a speech recognition processor. The position capture device corresponds to at least one trigger condition. When a sound source satisfies the trigger condition, the position acquisition device obtains the physical voice position of the sound source, and outputs the physical voice position. The directional sound pickup device is coupled to the position acquisition device, and the directional sound pickup device is used for receiving the physical voice position and receiving the voice signal generated by the sound source according to the physical voice position. The noise suppressor is coupled to the position acquisition device and the directional sound pickup device. The noise suppressor stores noise models corresponding to a plurality of speech generating positions respectively, and all the speech generating positions include physical speech positions. The noise suppressor is used for receiving the voice signal and the physical voice position, and eliminating the noise of the voice signal according to the noise model corresponding to the physical voice position, so as to generate a voice recognition signal. The speech recognition processor is coupled to the noise suppressor, wherein the speech recognition processor is used for receiving the speech recognition signal and generating an operation signal accordingly.

In an embodiment of the present invention, the speech recognition device further includes a coordinate converter, which is coupled to the position acquisition device, the noise suppressor and the directional sound pickup device. The coordinate converter is used to receive the physical voice position, and convert the coordinate system of the physical voice position to the coordinate system corresponding to the noise suppressor and the directional radio device, and then transmit the converted physical voice position to the noise suppressor and the directional radio device. .

In an embodiment of the present invention, the at least one position capture device includes a plurality of position capture devices, the at least one trigger condition includes a plurality of trigger conditions, and all the trigger conditions correspond to all the position capture devices respectively. When the audio source satisfies all trigger conditions in sequence, the position acquisition device corresponding to the earliest satisfied trigger condition obtains and outputs the physical voice position.

In an embodiment of the present invention, the position capturing device is an image positioning module. When the image positioning module captures an image with a gesture of raising a user's hand, the trigger condition is satisfied, the user is used as the audio source, and the physical position of the user is used as the physical voice position.

In an embodiment of the present invention, the position capturing device is a voice positioning module. When the voice positioning module receives the trigger voices generated by the audio source at different positions, the trigger condition is satisfied, and the voice positioning module is used to obtain different reception time points of the trigger voices at different positions, and obtain the physical voice position accordingly.

In an embodiment of the present invention, the position capture device includes a touch display panel and an application processor. The touch display panel is used to display the operation interface of the application program, wherein the operation interface has an image corresponding to the physical voice position. The application processor is coupled to the touch control display panel, the noise suppressor and the directional radio device, and the application processor is installed with an application program. When the position of the touch display panel corresponding to the image is pressed, the trigger condition is satisfied, and the application processor obtains and outputs the physical voice position.

In an embodiment of the present invention, the directional sound pickup device includes a microphone array and an audio processor. The microphone array is used to receive voice signals from different locations. The audio processor is coupled to the microphone array, the position acquisition device and the noise suppressor, and the audio processor stores a complex set of offset periods corresponding to all speech generating positions respectively. The audio processor is used to receive the physical voice position, and move the waveforms of the voice signals at different positions to the same time point according to the physical voice position and a set of offset time periods corresponding to the physical voice position, and add the voice signals at the same time point to generate the received voice signal. Enhanced voice signal. The audio processor is used to transmit the enhanced voice signal to the noise suppressor.

In one embodiment of the present invention, the directional sound pickup device includes a directional sound pickup and an automatic rotating platform. The directional microphone is coupled to the noise suppressor, the automatic rotating platform is coupled to the position capturing device, and the automatic rotating platform supports the directional microphone. The automatic rotating platform is used to receive the physical voice position and control the direction of the directional microphone to be directed towards the physical voice position. The directional radio is used to receive the voice signal and transmit the voice signal to the noise suppressor.

In an embodiment of the present invention, the speech recognition processor is coupled to the position acquisition device and the directional sound pickup device. When the voice recognition processor does not receive the voice recognition signal for a predetermined period of time, the voice recognition processor controls the position acquisition device to stop acquiring the physical voice position, and controls the directional radio device to stop receiving the physical voice position and generating the voice signal, and The control position acquisition device and the directional radio device operate in a standby state.

In one embodiment of the present invention, a speech recognition device is provided, which includes a plurality of speech receivers, an audio processor, a noise suppressor, and a speech recognition processor. All voice receivers are used to receive voice signals generated by an audio source at different locations. The audio processor is coupled to all the voice receivers, and the audio processor stores a complex set of offset periods corresponding to a plurality of voice generating positions respectively. The audio processor is used for obtaining different receiving time points of the voice signal at different positions, and obtaining the physical voice position of the sound source accordingly. All speech production locations contain physical speech locations. The audio processor is used for moving the waveforms of the voice signals at different positions to the same time point according to the physical voice position and a corresponding set of offset periods, and adding the voice signals at the same time point to generate the enhanced voice signal. The noise suppressor is coupled to the audio processor. The noise suppressor stores the noise models corresponding to all speech generating positions respectively. The noise suppressor is used for receiving the enhanced voice signal and the physical voice position, and eliminating the noise of the enhanced voice signal according to the noise model corresponding to the physical voice position, so as to generate a voice recognition signal. The speech recognition processor is coupled to the noise suppressor, wherein the speech recognition processor is used for receiving the speech recognition signal and generating an operation signal accordingly.

In an embodiment of the present invention, the speech recognition processor is coupled to the audio processor. When the voice recognition processor does not receive the voice recognition signal for a preset period of time, the voice recognition processor controls the audio processor to stop acquiring the physical voice position, and controls the audio processor to stop generating the enhanced voice signal, and controls the audio processor Operation is in standby mode.

Based on the above, the voice recognition device first obtains the physical voice position of the audio source, and outputs the voice position to the directional radio device, so that the directional radio device receives the voice signal generated by the audio source according to the voice position. In this way, when the voice signal controls the public equipment, the frequency of grabbing the control rights is reduced and the operability of the public equipment is improved, and in the complex and closed environment, the sound quality, the direction of the sound and the noise reduction function are improved to improve the voice. Accuracy of identification.

Hereby, in order to make your examiners have a further understanding and understanding of the structural features of the present invention and the effects achieved, I would like to assist with the preferred embodiment drawings and coordinate detailed descriptions, and the descriptions are as follows:

Embodiments of the present invention will be further explained with the help of the related drawings below. Wherever possible, in the drawings and the description, the same reference numbers refer to the same or similar components. In the drawings, shapes and thicknesses may be exaggerated for simplicity and convenience. It should be understood that the elements not particularly shown in the drawings or described in the specification have forms known to those of ordinary skill in the art. Those skilled in the art can make various changes and modifications based on the content of the present invention.

The disclosure is specifically described with the following examples, which are only for illustration, because for those skilled in the art, various changes and modifications can be made without departing from the spirit and scope of the present disclosure. The scope of protection of the disclosed contents shall be determined by the scope of the appended patent application. Throughout the specification and claims, unless the content clearly dictates otherwise, the meanings of "a" and "the" include that such recitations include "one or at least one" of the element or ingredient. Furthermore, as used in this disclosure, a singular article also includes the recitation of a plurality of elements or components unless the exclusion of the plural is obvious from the specific context. Also, as used in this description and throughout the claims below, the meaning of "in" may include "in" and "on" unless the content clearly dictates otherwise. Terms used throughout the specification and the scope of the patent application, unless otherwise specified, generally have the ordinary meaning of each term used in the field, in the content disclosed herein and in the specific content. Certain terms used to describe the present disclosure are discussed below or elsewhere in this specification to provide practitioners with additional guidance in describing the present disclosure. Examples anywhere throughout the specification, including the use of examples of any terms discussed herein, are by way of illustration only, and of course do not limit the scope and meaning of the disclosure or any exemplified terms. Likewise, the present disclosure is not limited to the various embodiments set forth in this specification.

Furthermore, if the term "electrically (sexually) coupled" or "electrically (sexually) connected" is used herein, it includes any means of direct and indirect electrical connection. For example, if it is described in the text that a first device is electrically coupled to a second device, it means that the first device can be directly connected to the second device, or indirectly connected to the second device through other devices or connecting means device. In addition, if the transmission and provision of electrical signals are described, those skilled in the art should understand that the transmission of electrical signals may be accompanied by attenuation or other non-ideal changes. In fact, it should be regarded as the same signal. For example, if the electrical signal S is transmitted (or provided) from the terminal A of the electronic circuit to the terminal B of the electronic circuit, it may pass through the source and drain terminals of a transistor switch and/or possible stray capacitance. A voltage drop is generated, but the purpose of this design is not to deliberately use the attenuation or other non-ideal changes generated during transmission (or supply) to achieve some specific technical effects. The electrical signal S is at the terminal A and the terminal of the electronic circuit. B should be regarded as substantially the same signal.

The following description of "one embodiment" or "an embodiment" refers to a particular element, structure or feature associated with at least one embodiment. Thus, the appearances of "one embodiment" or "an embodiment" in various places below are not directed to the same embodiment. Furthermore, the specific components, structures and features in one or more embodiments may be combined in a suitable manner.

Unless otherwise specified, some conditional sentences or words, such as "can", "could", "might", or "may", are usually intended to express that the embodiments of this case have, However, it can also be interpreted as features, elements, or steps that may not be required. In other embodiments, these features, elements, or steps may not be required.

FIG. 1 is a circuit block diagram of a speech recognition apparatus according to a first embodiment of the present invention. Referring to FIG. 1, the following describes the first embodiment of the speech recognition device of the present invention. The speech recognition device 1 includes at least a position acquisition device 10 , a directional sound pickup device 11 , a noise suppressor 12 and a speech recognition processor 13 , wherein these components are all hardware. The directional sound pickup device 11 is coupled to the position capture device 10 , the noise suppressor 12 is coupled to the position capture device 10 and the directional sound pickup device 11 , and the speech recognition processor 13 is coupled to the noise suppressor 12 . At least one position capture device 10 corresponds to at least one trigger condition. For clarity and convenience, the number of the position capturing device 10 and the triggering condition is taken as an example. In addition, the position acquisition device 10 , the directional sound pickup device 11 and the noise suppressor 12 can use the same coordinate system.

The operation process of the first embodiment is described below. When a sound source 2 satisfies the trigger condition, the position capturing device 10 obtains the physical voice position P of the sound source 2 and outputs the physical voice position P. The directional sound pickup device 11 receives the physical voice position P, and receives the voice signal V generated by the audio source 2 according to the physical voice position P, wherein the voice signal V includes the operation voice corresponding to the operation right. For example, the directional radio device 11 can be implemented by a beamforming module, so as to strengthen the voice signal V in the direction corresponding to the physical voice position P, and weaken the voice signal V in other directions. Since the noise suppressor 12 has a plurality of noise models corresponding to the speech generating positions respectively, all the speech generating positions include the physical speech position P. As shown in FIG. Therefore, the noise suppressor 12 receives the voice signal V and the physical voice position P, and cancels the noise of the voice signal V according to the noise model corresponding to the physical voice position P to generate a voice recognition signal R. The noise suppressor 12 can further use an adaptive filter algorithm and a finite impulse response (Finite impulse response, FIR) filter to eliminate the noise of the voice signal V, so as to improve the noise suppression efficiency. The speech recognition processor 13 receives the speech recognition signal R, and generates an operation signal O accordingly. The operation signal O can be used to control the public equipment. The position acquisition device 10 first obtains the physical voice position P of the sound source 2 and outputs the voice position P to the directional sound pickup device 11 , so that the directional sound pickup device 11 receives the voice signal V generated by the sound source 2 according to the physical voice position P. In this way, when the voice signal V controls the public equipment, the frequency of grabbing control rights can be reduced and the operability of the public equipment can be improved. Improve the accuracy of speech recognition.

In some embodiments of the present invention, the speech recognition processor 13 may be coupled to the position capturing device 10 and the directional sound pickup device 11 . When the speech recognition processor 13 does not receive the speech recognition signal R for a predetermined period of time, it means that the operation of the speech recognition device 1 is finished, so as to release the operation right. When the speech recognition processor 13 does not receive the speech recognition signal R within the preset time period, the speech recognition processor 13 controls the position capturing device 10 to stop acquiring the physical voice position P, and controls the directional sound pickup device 11 to stop receiving the physical voice position P Then, a voice signal V is generated, and the position capturing device 10 and the directional sound receiving device 11 are controlled to operate in a standby state until the position capturing device 10 captures a new physical position of a new audio source.

FIG. 2 is a circuit block diagram of a speech recognition device according to a second embodiment of the present invention. Please refer to FIG. 2 , the second embodiment of the speech recognition device of the present invention will be described below. The difference between the second embodiment and the first embodiment lies in the number of position capturing devices 10 and triggering conditions thereof. In the second embodiment, there are a plurality of position capturing apparatuses 10 and a plurality of trigger conditions. In order to prevent the voice signal V generated by the audio source 2 from being masked and unable to satisfy a single trigger condition, the second embodiment uses different trigger conditions, such as voice-related trigger conditions, image-related trigger conditions, and application-related trigger conditions. All trigger conditions correspond to all position capture devices 10 respectively. The present invention does not consider the state in which multiple trigger conditions are simultaneously triggered. When the audio source 2 satisfies all trigger conditions in sequence, the physical voice position P is obtained and output by the position acquisition device 10 corresponding to the earliest satisfied trigger condition.

FIG. 3 is a circuit block diagram of a speech recognition apparatus according to a third embodiment of the present invention. Referring to FIG. 3, the following describes the third embodiment of the speech recognition device of the present invention. The difference between the third embodiment and the first embodiment is that the third embodiment further includes a coordinate converter 14 . In the third embodiment, the position acquisition device 10 and the directional sound pickup device 11 can use different coordinate systems, and the directional sound pickup device 11 and the noise suppressor 12 can use the same coordinate system. The coordinate converter 14 is coupled to the position acquisition device 10 , the noise suppressor 12 and the directional sound pickup device 11 . The coordinate converter 14 receives the physical voice position P, and converts the coordinate system of the physical voice position P to the coordinate system corresponding to the noise suppressor 12 and the directional radio device 11, and then transmits the converted physical voice position P' to the noise suppressor The device 12 and the directional sound pickup device 11, in which all speech generating positions also include the converted physical speech position P'. Therefore, the directional sound pickup device 11 receives the converted physical voice position P', and receives the voice signal V generated by the audio source 2 according to the converted physical voice position P'. The noise suppressor 12 receives the voice signal V and the converted physical voice position P', and eliminates the noise of the voice signal V according to the noise model corresponding to the converted physical voice position P', so as to generate a voice recognition signal R .

In one embodiment of the present invention, the position capturing device 10 may be an image positioning module, and the trigger condition is an image-related trigger condition. When the image positioning module captures an image with a specific gesture of a user, such as a raised hand gesture, the trigger condition is satisfied, the user is the audio source 2 , and the physical position of the user is the physical voice position P. For example, the image positioning module can divide the captured image into a plurality of blocks, and mark each block with a number, so that the number of the block with the raised hand gesture can be known and used as Entity speech position P. Alternatively, if the image positioning module has dual lenses, the image positioning module can use the dual lenses to position the user, so as to obtain the three-dimensional coordinates of the user, and use this as the physical voice position P.

In another embodiment of the present invention, the position capturing device 10 may be a voice positioning module, and the triggering condition is a voice-related triggering condition. When the voice location module receives the trigger voice generated by the audio source 2 at different positions, the trigger condition is satisfied. The trigger voice can be the same as or different from the voice signal. The voice positioning module obtains different reception time points of the trigger voice at different positions. Since the different receiving time points respectively represent the distances between the audio source 2 and the different positions of the voice positioning module, the voice positioning module can obtain the physical voice position P according to different receiving time points. For example, the speech positioning module may include a stereo microphone array and a speech processor coupled to each other. The stereo microphone array includes a plurality of microphones. Because all the microphones are located at different positions, all the microphones will receive the sound source at different time points. For the trigger voice generated by 2, the voice processor can calculate the three-dimensional coordinates of the sound source 2 according to the time interval of different time points and the positions of all microphones, and use this as the physical voice position P.

FIG. 4 is a circuit block diagram of the position acquisition device 10 , the directional sound pickup device 11 and the noise suppressor 12 according to an embodiment of the present invention. Referring to FIG. 4 , the position capture device 10 may include a touch display panel 100 and an application processor 101 , wherein the application processor 101 is coupled to the touch display panel 100 , the noise suppressor 12 and the directional sound pickup device 11 . The touch display panel 100 displays an operation interface of the application program, wherein the operation interface has an image corresponding to the physical voice position P. The application processor 101 has an application program installed, so the trigger condition is an application program-related trigger condition. When the position of the touch display panel 100 corresponding to the above-mentioned image is pressed, the trigger condition is satisfied, and the application processor 101 obtains and outputs the physical voice position P. Furthermore, the circuit shown in Fig. 4 may be applied to Fig. 1 or other embodiments of the present invention, but is not limited thereto. When the circuit shown in FIG. 4 is applied in the embodiment of FIG. 3 , the application processor 101 is coupled to the coordinate converter 14 .

FIG. 5 is a circuit block diagram of the position acquisition device 10 , the directional sound pickup device 11 and the noise suppressor 12 according to another embodiment of the present invention. Please refer to FIG. 5 , the directional sound pickup device 11 may include a microphone array 110 and an audio processor 111 . The microphone array 110 receives the voice signals V at different positions. The audio processor 111 is coupled to the microphone array 110 , the position acquisition device 10 and the noise suppressor 12 , and the audio processor 111 stores a complex array of offset periods corresponding to all speech generating positions respectively. The audio processor 111 receives the physical voice position P, moves the waveforms of the voice signals V at different positions to the same time point according to the physical voice position P and a set of offset periods corresponding to the physical voice position P, and adds the voice signal V at the same time point , to generate an enhanced voice signal V'. The audio processor 111 transmits the enhanced voice signal V' to the noise suppressor 12, so that the noise suppressor 12 eliminates the noise of the enhanced voice signal V' according to the noise model corresponding to the physical voice position P, so as to generate a Voice recognition signal R. Furthermore, the circuit shown in Fig. 5 may be applied to Fig. 1 or other embodiments of the present invention, but is not limited thereto. When the circuit shown in FIG. 5 is applied in the embodiment of FIG. 3, the audio processor 111 and the noise suppressor 12 are coupled to the coordinate converter 14, and the converted physical speech position P' replaces the physical speech position p. When the circuit shown in FIG. 5 is applied in the embodiment of FIG. 4 , the audio processor 111 is coupled to the application processor 101 .

FIG. 6 is a schematic circuit diagram of a sound source 2 and a directional sound pickup device 11 according to an embodiment of the present invention. Referring to FIG. 6, the microphone array 110 may include microphones m1, m2 and m3, the audio processor 111 may include time shifters 1111, 1111' and 1111", an average calculator 1112 and a parameter adjuster 1113, wherein the time The shifters 1111, 1111' and 1111" are respectively coupled to the microphones m1, m2 and m3, the parameter adjuster 1113 is coupled to the time shifters 1111, 1111' and 1111", and the time shifters 1111, 1111' and 1111" are coupled Connect to Average Calculator 1112. The parameter adjuster 1113 stores the complex group offset periods corresponding to all speech generating positions respectively. Because the distances between the microphones m1, m2, and m3 from the sound source 2 are all different, the microphones m1, m2, and m3 receive the voice signal V at different time points. For example, the time interval between the voice signals V received by the microphones m2 and m3 is t1, and the time interval between the voice signals V received by the microphones m1 and m3 is t2. It is assumed that the converted physical voice position P' or the physical voice position P corresponds to the microphone m3, which means that the microphone m3 is the closest to the sound source 2. The parameter adjuster 1113 adjusts the offset periods of the time offsetrs 1111, 1111' and 1111" to be d1, d2 and d3, respectively, so that d1=t2, d2=t1, d3=0. Therefore, the microphones m1, m2 and m3 The waveform of the received voice signal V is shifted to the corresponding time point when the microphone m3 receives the voice signal V. Next, the average calculator 1112 receives all the voice signals V from the time shifters 1111, 1111' and 1111", They are summed and averaged to generate the enhanced speech signal V'. Furthermore, the circuit shown in Fig. 6 may be applied to Fig. 1 or other embodiments of the present invention, but is not limited thereto. When the circuit shown in FIG. 6 is applied in the embodiment of FIG. 1 , the parameter adjuster 1113 is coupled to the position capturing device 10 . When the circuit shown in FIG. 6 is applied in the embodiment of FIG. 3 , the parameter adjuster 1113 is coupled to the coordinate converter 14 . When the circuit shown in FIG. 6 is applied in the embodiment of FIG. 4 , the parameter adjuster 1113 is coupled to the application processor 101 .

FIG. 7 is a circuit block diagram of a position acquisition device 10 , a directional sound pickup device 11 and a noise suppressor 12 according to still another embodiment of the present invention. Please refer to FIG. 7 , the directional sound pickup device 11 may also include a directional sound pickup device 112 and an automatic rotating platform 113 . The directional microphone 112 is coupled to the noise suppressor 12 , the automatic rotating platform 113 is coupled to the position capturing device 10 , and the automatic rotating platform 113 supports the directional microphone 112 . The automatic rotating platform 113 receives the physical voice position P, and controls the direction of the directional microphone 112 to be directed toward the physical voice position P, and the directional microphone 112 receives the voice signal V and transmits the voice signal V to the noise suppressor 12 . In addition, the circuit shown in FIG. 7 may be applied to other embodiments in FIG. 1 or the present invention, but is not limited thereto. When the circuit shown in FIG. 7 is applied in the embodiment of FIG. 3, the automatic rotating platform 113 and the noise suppressor 12 are coupled to the coordinate converter 14, and the converted physical speech position P' is used to replace the physical speech position p. When the circuit shown in FIG. 7 is applied in the embodiment of FIG. 4 , the automatic rotation platform 113 is coupled to the application processor 101 .

FIG. 8 is a circuit block diagram of a speech recognition apparatus according to a fourth embodiment of the present invention. Referring to FIG. 8 , the fourth embodiment of the speech recognition apparatus of the present invention is described below. The voice recognition device 3 includes a plurality of voice receivers 30 , an audio processor 31 , a noise suppressor 32 and a voice recognition processor 33 , all of which are hardware. The audio processor 31 is coupled to all the voice receivers 30, and the audio processor 31 stores a plurality of offset periods corresponding to the voice generating positions respectively. The noise suppressor 32 is coupled to the audio processor 31 , wherein the noise suppressor 32 stores the noise models corresponding to all speech generating positions respectively. The speech recognition processor 33 is coupled to the noise suppressor 32 . In addition, the audio processor 31 and the noise suppressor 32 use the same coordinate system.

The operation of the fourth embodiment is described below. First, all the voice receivers 30 receive a voice signal V generated by an audio source 4 at different positions, wherein the voice signal V includes the operation voice corresponding to the operation right. Because the distances between all the voice receivers 30 and the sound source 4 are different, all the voice receivers 30 will receive the voice signal V at different time points. The audio processor 31 obtains different reception time points of the voice signal V at different positions, and obtains the physical voice position P of the audio source 4 accordingly, and all the voice generating positions include the physical voice position P. The audio processor 31 moves the waveforms of the voice signals V at different positions to the same time point according to the physical voice position P and its corresponding set of offset periods, and adds the voice signals V at the same time point to generate an enhanced voice signal V'. The noise suppressor 32 receives the enhanced voice signal V' and the physical voice position P, and eliminates the noise of the enhanced voice signal V' according to the noise model corresponding to the physical voice position P to generate a voice recognition signal R. The noise suppressor 32 can further use an adaptive filter algorithm and a Finite Impulse Response (FIR) filter to eliminate the noise of the voice signal V, so as to improve the noise suppression efficiency. The speech recognition processor 33 receives the speech recognition signal R, and generates an operation signal O accordingly. The operation signal O can be used to control the public equipment. The audio processor 31 first obtains the physical voice position P of the audio source 4, so that the audio processor 31 generates an enhanced voice signal V' according to the physical voice position P. In this way, when the voice signal V controls the public equipment, the frequency of grabbing control rights can be reduced and the operability of the public equipment can be improved. Improve the accuracy of speech recognition.

In some embodiments of the present invention, the speech recognition processor 33 may be coupled to the audio processor 31 . When the speech recognition processor 33 does not receive the speech recognition signal R for a predetermined period of time, it means that the operation of the speech recognition device 3 is ended, so as to release the operation right. When the voice recognition processor 33 does not receive the voice recognition signal R for a preset period of time, the voice recognition processor 33 controls the audio processor 31 to stop acquiring the physical voice position P, and controls the audio processor 31 to stop generating the enhanced voice signal V ', and control the audio processor 31 to operate in a standby state until the voice receiver 30 receives a new voice signal.

FIG. 9 is a schematic circuit diagram of an audio source 4 , a voice receiver 30 and an audio processor 31 according to an embodiment of the present invention. Please refer to FIG. 8 and FIG. 9, the voice receiver 30 may be implemented by microphones M1, M2 and M3, and the audio processor 31 may include time shifters 311, 311' and 311", an average calculator 312 and a parameter adjustment 313, wherein the time shifters 311, 311' and 311" are respectively coupled to the microphones M1, M2 and M3, and the parameter adjuster 313 is coupled to the time shifters 311, 311' and 311" and the microphones M1, M2 and M3, The time shifters 311 , 311 ′ and 311 ″ are coupled to the average calculator 312 . The parameter adjuster 313 stores the complex group offset periods corresponding to all speech generating positions respectively. Because the distances between the microphones M1, M2 and M3 from the sound source 4 are all different, the microphones M1, M2 and M3 will receive the voice signal V at different time points. For example, the time interval between the voice signals V received by the microphones M2 and M3 is t1, and the time interval between the voice signals V received by the microphones M1 and M3 is t2. It is assumed that the parameter adjuster 313 finds that the physical voice position P corresponds to the microphone M3 , which means that the microphone M3 is the closest to the sound source 4 . The parameter adjuster 313 adjusts the offset periods of the time offsetrs 311, 311' and 311" to be d1, d2 and d3, respectively, so that d1=t2, d2=t1, and d3=0. Therefore, the microphones M1, M2 and M3 The waveform of the received voice signal V is shifted to the corresponding time point when the microphone M3 receives the voice signal V. Next, the average calculator 312 receives all the voice signals V from the time shifters 311, 311' and 311", They are summed and averaged to generate the enhanced speech signal V'. In addition, the circuit shown in FIG. 9 may be applied to FIG. 8 or other embodiments of the present invention, but is not limited thereto.

According to the above-mentioned embodiment, the voice recognition device first obtains the physical voice position of the audio source, and outputs the voice position to the directional radio device, so that the directional radio device receives the voice signal generated by the audio source according to the voice position. In this way, when the voice signal controls the public equipment, the frequency of grabbing the control rights is reduced and the operability of the public equipment is improved, and in the complex and closed environment, the sound quality, the direction of the sound and the noise reduction function are improved to improve the voice. Accuracy of identification.

The above description is only a preferred embodiment of the present invention, and is not intended to limit the scope of the present invention. Therefore, all changes and modifications made in accordance with the shape, structure, feature and spirit described in the scope of the patent application of the present invention are equivalent. , shall be included in the scope of the patent application of the present invention.

1: Voice recognition device 10: Location capture device 100: Touch display panel 101: Application Processors 11: Directional radio device 110: Microphone Array 111: Audio Processor 1111, 1111', 1111": Time Offset 1112: Average Calculator 1113: Parameter adjuster 112: Directional Radio 113: Automatic rotating platform 12: Noise suppressor 13: Speech recognition processor 14: Coordinate converter 2: Audio source 3: Voice recognition device 30: Voice Receiver 31: Audio Processor 311, 311', 311": Time Offset 312: Average Calculator 313: Parameter adjuster 32: Noise suppressor 33: Speech recognition processor 4: Audio source P: Physical voice position V: voice signal R: voice recognition signal O: Operation signal P': the transformed physical voice position V’: Enhanced voice signal m1, m2, m3: microphone M1, M2, M3: Microphone

FIG. 1 is a circuit block diagram of a speech recognition apparatus according to a first embodiment of the present invention. FIG. 2 is a circuit block diagram of a speech recognition device according to a second embodiment of the present invention. FIG. 3 is a circuit block diagram of a speech recognition apparatus according to a third embodiment of the present invention. FIG. 4 is a circuit block diagram of a position acquisition device, a directional radio device and a noise suppressor according to an embodiment of the present invention. FIG. 5 is a circuit block diagram of a position acquisition device, a directional radio device and a noise suppressor according to another embodiment of the present invention. FIG. 6 is a schematic circuit diagram of a sound source and a directional sound pickup device according to an embodiment of the present invention. FIG. 7 is a circuit block diagram of a position acquisition device, a directional radio device and a noise suppressor according to still another embodiment of the present invention. FIG. 8 is a circuit block diagram of a speech recognition apparatus according to a fourth embodiment of the present invention. FIG. 9 is a schematic circuit diagram of an audio source, a voice receiver and an audio processor according to an embodiment of the present invention.

1: Voice recognition device

10: Location capture device

11: Directional radio device

12: Noise suppressor

13: Speech recognition processor

2: Audio source

P: Physical voice position

V: voice signal

R: voice recognition signal

O: Operation signal

Claims (9)

A voice recognition device, comprising: at least one position capture device corresponding to at least one trigger condition, when a sound source satisfies the at least one trigger condition, the at least one position capture device obtains the physical voice position of the sound source, and outputs the a physical voice position, wherein the at least one trigger condition is a voice-related trigger condition, an image-related trigger condition or an application-related trigger condition; a directional sound pickup device coupled to the at least one position capture device, the directional sound pickup device is used for Receive the physical voice position, and receive the voice signal generated by the audio source according to the physical voice position; a noise suppressor, coupled to the at least one position acquisition device and the directional sound pickup device, wherein the noise suppressor stores A noise model corresponding to a plurality of voice generating positions respectively, the voice generating positions include the physical voice position, the noise suppressor is used for receiving the voice signal and the physical voice position, and according to the noise corresponding to the physical voice position The model eliminates the noise of the voice signal to generate a voice recognition signal; and a voice recognition processor is coupled to the noise suppressor, wherein the voice recognition processor is used to receive the voice recognition signal and generate an operation accordingly signal; wherein the speech recognition processor is coupled to the at least one position acquisition device and the directional sound pickup device, and when the speech recognition processor does not receive the speech recognition signal for a preset period of time, the speech recognition processor controls The at least one position capturing device stops acquiring the physical voice position, and controls the directional sound pickup device to stop receiving the physical voice position and generating the voice signal, and controls the at least one position capturing device and the directional sound pickup device to operate in Standby state. The speech recognition device as claimed in claim 1, further comprising a coordinate converter, which is coupled to the at least one position acquisition device, the noise suppressor and the directional radio device, wherein the coordinate converter is used to receive the The physical voice position, and the coordinate system of the physical voice position is converted into the coordinate system corresponding to the noise suppressor and the directional radio device, and then the converted physical voice position is sent to the noise suppressor and the directional radio. device. The speech recognition device as claimed in claim 1, wherein the at least one position capture device includes a plurality of position capture devices, the at least one trigger condition includes a plurality of trigger conditions, and the trigger conditions respectively correspond to the position capture devices , when the audio source sequentially satisfies the triggering conditions, the position acquisition device corresponding to the triggering condition that is satisfied earliest obtains and outputs the physical voice position. The speech recognition device of claim 1, wherein the at least one position capturing device is an image positioning module, and when the image positioning module captures an image with a user's raised hand gesture, the at least one trigger condition is satisfied, the user is the audio source, and the physical location of the user is the physical voice location. The voice recognition device according to claim 1, wherein the at least one position capturing device is a voice positioning module, and the at least one triggering condition is satisfied when the voice positioning module receives trigger voices generated by the audio source at different positions , and the voice positioning module is used to obtain different reception time points of the trigger voice at the different positions, and obtain the physical voice position accordingly. The speech recognition device of claim 1, wherein the at least one location captures The retrieval device includes: a touch display panel for displaying an operation interface of an application program, wherein the operation interface has an image corresponding to the physical voice position; and an application processor, coupled to the touch display panel, the noise suppression a device and the directional radio device, the application processor is installed with the application program, wherein when the touch display panel is pressed at the position corresponding to the image, the at least one trigger condition is satisfied, and the application processor obtains and outputs The entity's voice location. The voice recognition device of claim 1, wherein the directional sound pickup device comprises: a microphone array for receiving the voice signals at different positions; and an audio processor coupled to the microphone array, the at least one position capture Taking the device and the noise suppressor, the audio processor stores a complex set of offset periods corresponding to the voice generating positions respectively, wherein the audio processor is used for receiving the physical voice position and corresponding to the physical voice position according to the physical voice position. The set of offset periods shifts the waveforms of the voice signals at the different positions to the same time point, and adds the voice signals at the same time point to generate the enhanced voice signal, and the audio processor is used to transmit the voice signal. The voice signal is enhanced to the noise suppressor. The speech recognition device of claim 1, wherein the directional sound pickup device comprises: a directional sound receiver coupled to the noise suppressor; and an automatic rotating platform coupled to the at least one position acquisition device, the The automatic rotating platform supports the directional microphone, wherein the automatic rotating platform is used to receive the voice position of the entity and control the sound-receiving direction of the directional microphone toward the entity voice position, and the directional microphone is used for receiving the voice signal and transmitting the voice signal to the noise suppressor. A voice recognition device, comprising: a plurality of voice receivers for receiving voice signals generated by an audio source at different positions; an audio processor, coupled to the voice receivers, the audio processor stores a plurality of voice generation positions A complex set of offset periods corresponding respectively, the audio processor is used to obtain different reception time points of the voice signal at the different positions, and obtain the physical voice positions of the audio source accordingly, and the voice production positions include the physical voice positions , wherein the audio processor is used to move the waveforms of the voice signals at the different positions to the same time point according to the physical voice position and the corresponding set of offset periods, and add the voice signals at the same time point to generate the enhanced voice signal; a noise suppressor coupled to the audio processor, wherein the noise suppressor stores the noise models corresponding to the voice generating positions respectively, and the noise suppressor is used for receiving the enhanced voice signal the voice signal and the physical voice position, and eliminate the noise of the enhanced voice signal according to the noise model corresponding to the physical voice position to generate a voice recognition signal; and a voice recognition processor, coupled to The noise suppressor, wherein the voice recognition processor is used for receiving the voice recognition signal, and generates an operation signal accordingly; wherein the voice recognition processor is coupled to the audio processor, and the voice recognition processor does not receive the voice When the recognition signal lasts for a predetermined period of time, the voice recognition processor controls the audio processor to stop acquiring the physical voice position, controls the audio processor to stop generating the enhanced voice signal, and controls the audio processor to stop generating the enhanced voice signal. The audio processor is operating in a standby state.

Images