JP2008048342A - Sound acquisition apparatus - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a sound acquisition apparatus which securely converts the speaking speed of the speech of only a speaking person existing at an optional position surrounding the apparatus from acquired speeches but never converts the speaking speed of a background sound. <P>SOLUTION: A speech signal processing part 4 imparts prescribed delay to speech signals acquired by each microphone 2, and forms a sound acquisition beam surrounding the microphone 2. A controller 8 generates information showing the existing area of the speaking person (speaking person position information) based on an area corresponding to the sound acquisition beam of the highest level, and outputs the information to a storage part 3 to be recorded. The sound acquisition beam corresponding to the speaking person position information is output to a speaking speed conversion part 5 as a speaking sound voice signal, and the other sound acquisition beams are output to a mixer 6 as background speech signals. Thus, while the speaking speed of the speech of only the speaking person is converted but the speaking speed of the background sound is not converted, the speech can be released and recorded. <P>COPYRIGHT: (C)2008,JPO&INPIT

Description

  The present invention relates to a sound collection device that is used in a conference and the like and collects speech sounds of conference participants.

  2. Description of the Related Art Conventionally, there has been proposed an apparatus that makes it easy to listen to the content of an utterance by expanding an input voice signal on the time axis and converting the speech speed. However, when the input voice signal is expanded, sounds other than the speaker's voice (for example, BGM) are also expanded at the same time. Further, the BGM is expanded even when the voice of the speaker is not input. When the listener is listening to the BGM at the same time (in parallel) with the speaker's voice, if the listener is extended to the BGM, there is a problem that the atmosphere of the original musical sound cannot be felt.

  In view of this, an apparatus has been proposed in which speech speed conversion processing is performed only when an input speech signal is analyzed and determined to be speaker speech (see, for example, Patent Document 1).

  In addition, a device has been proposed in which a plurality of microphones are installed, and the collected voices from the same distance from each microphone (in the same phase) are used as speech voices and other collected voices are separated as background sounds (for example, Patent Document 2).

There has also been proposed an apparatus configured to handle voice and background sounds with a plurality of independent channels and perform speech speed conversion processing only on the voice channels (see, for example, Patent Document 3).
JP 2000-152394 A JP 2005-208173 A JP 2004-244081 A

  However, the apparatus of Patent Document 1 has a problem that the background sound collected at the same timing as the uttered voice is converted into the speech speed in the same manner as the uttered voice.

  Further, since the apparatus of Patent Document 2 can process only speech from a point where the distance from each microphone is equal as speech speech, if there is a speaker other than this point, speech speed cannot be converted for that speaker's speech. There was a problem.

  Further, in the apparatus of Patent Document 3, it is necessary to record the speech sound and the background sound in separate channels when recording, and the speaker needs to speak to the microphone assigned to the specific channel.

  It is an object of the present invention to provide a sound collection device that accurately converts only the voice of a speaker existing at an arbitrary position around the device from the collected sound and does not convert the background sound. .

The sound collection device of the present invention includes a microphone array in which a plurality of microphones are arranged,
A sound collection control unit that forms a sound collection beam for a plurality of user directions and compares the sound collection beam intensities to identify a speaker direction, and uses the sound collection beam of the speaker direction as a speech signal A speech signal selection means for selecting a sound collection beam other than the sound collection beam of the speaker direction as a background voice signal, a speech speed conversion means for converting the speech speed of the speech voice signal, and the speech speed conversion means. And a mixer that mixes the converted speech audio signal and the background audio signal selected by the audio signal selection means.

  In the present invention, a predetermined delay is given to the collected sound signal of each microphone, and a plurality of sound collecting beams having strong directivity in a specific direction are formed. The speaker orientation is identified by comparing the levels of these sound collecting beams. For example, the direction of the sound collecting beam having the highest level is set as the speaker orientation. The collected sound beam in the direction of the speaker is used as a speaker voice signal, which is converted into the speech speed and then output to the mixer. The collected sound beam in the other direction is output to the mixer as it is without converting the speech speed.

  In the sound collecting device of the present invention, when there is a sound collecting beam of a predetermined level or more in a direction other than the sound collecting beam selected as the speech sound signal, the sound signal selecting means has a sound collecting beam in that direction. Is selected as a background audio signal.

  In the present invention, when a high-level sound collecting beam exists in a direction other than the direction in which it is determined that a speaker is present, it is assumed that a sound source of background sound exists in that direction, and the sound collecting beam in that direction is input to the mixer as a background sound signal Output. Thereby, it is possible to accurately collect the background voice.

  Further, in the sound collecting device of the present invention, the sound signal selecting means includes a sound collecting beam selected as the uttered sound signal, and a sound collecting beam in a direction adjacent to the sound collecting beam selected as the uttered sound signal. The difference signal is input to the speech speed conversion means as an utterance voice signal.

  In the present invention, the sound collecting beam in the adjacent direction is subtracted from the sound collecting beam selected as the speaker voice signal. Thereby, the level of the background voice included in the collected sound beam selected as the speaker voice signal can be reduced, and only the voice of the speaker can be converted more accurately.

  The sound collection device according to the present invention further includes speech sound signal extraction means for extracting speech signals of speech sound from a plurality of sound collection beams formed by the sound collection control unit, and the sound collection control unit includes a plurality of sound collection control units. It is characterized in that the direction of the sound collecting beam having the highest level among the sound collecting beams and from which the speech signal extracting means extracts the speech signal of the speech is determined as the speaker orientation.

  In the present invention, the speech signal of the speech speech is extracted from each sound collection beam. For example, the voice feature amount of the collected sound beam is extracted, compared with the voice feature amount of the uttered voice stored in advance, and if they match, the voice is estimated. The sound collection control unit selects the sound collection beam having the highest level and the sound signal estimated as the speech sound as the speaker sound signal, so that only the sound of the speaker can be converted more accurately. Can do.

  According to the present invention, the direction of the speaker is determined by the sound collection beam formed by the microphone array, the speech speed is converted only for the sound collection beam with respect to the direction of the speaker, and the sound collection beam in the other direction is output as it is. Thus, only the voice of the speaker can be accurately converted, and the background sound can be collected without converting the voice speed.

  With reference to the drawings, a sound emission and collection device according to an embodiment of the present invention will be described. This sound emission and collection device is used as a loudspeaker, a sound recorder, or the like in a conference. FIG. 1 is a block diagram showing a configuration of a sound emission and collection device. As shown in the figure, the sound emission and collection device includes a speaker 1, a plurality of microphones 2A to 2M, a storage unit 3, an audio signal processing unit 4, a speech rate conversion unit 5, a mixer 6, a recording / playback unit 7, and a controller. 8 and an input / output I / F 9.

  The plurality of microphones 2A to 2M are arranged in a straight line (or matrix shape or honeycomb shape) at regular intervals to constitute a microphone array. Each microphone 2 is generally a dynamic microphone, but other types such as a condenser microphone may be used. The number of microphones and the interval between the microphones are appropriately set according to the environment where the sound emitting and collecting apparatus is installed, the required frequency band, and the like.

  When sound is emitted at a certain position around the microphones 2A to 2M, each microphone 2 picks up the sound. The microphone 2 outputs an audio signal from the collected audio to the audio signal processing unit 4. In FIG. 1, front-end amplifiers and A / D converters for converting analog audio signals into digital audio signals are omitted. The audio signal output from each microphone 2 is synthesized by the audio signal processing unit 4 and output to the speech speed conversion unit 5 or the mixer 6. The audio signal processing unit 4 selectively outputs the audio signal output from each microphone 2 in accordance with an instruction from the controller 8. When sound is collected by each microphone 2, the sound is propagated with a propagation time corresponding to the distance between each microphone 2 and the sound source, so that each microphone 2 has a time difference in sound collection timing.

  Here, for example, if sound waves arrive at all the microphones 2 from the front at the same timing, the audio signals output from the respective microphones 2 are strengthened by synthesis. On the other hand, when sound waves arrive from other directions, the audio signals output from the microphones 2 are weakened by being synthesized because they have different phases. Therefore, the sensitivity of the array microphone is narrowed down into a beam shape, and the main sensitivity (sound collecting beam) is formed only in the front.

  The audio signal processing unit 4 can direct the sound collection beam obliquely by giving a predetermined delay time to the audio signal output from each microphone 2. When the sound collecting beam is inclined, the sound signal is set to be sequentially output from the adjacent microphone 2 every time a predetermined time elapses from one end microphone 2. For example, when the sound source is present in front of one end of the microphone array, the sound wave comes from one end closest to the sound source, and the sound wave comes last to the opposite end. In order to correct this propagation time difference, the audio signal of each microphone 2 is added with a delay time and then synthesized. Thereby, the audio signal in this direction can be strengthened by synthesis. Therefore, by sequentially delaying the audio signals output from the microphones 2 arranged in a row from one end to the other end, the sound collection beam is inclined according to the delay time.

  Also, a plurality of sound collecting beams can be formed simultaneously. FIG. 2 is a block diagram illustrating a configuration of a main part connected to the microphone 2 in the audio signal processing unit 4. The microphones 2A to 2M are connected to the digital filters 41A to 41M of the audio signal processing unit 4, respectively. The sound collected by the microphones 2A to 2M is input to the digital filters 41A to 41M as a digital sound signal. 2 shows a detailed block diagram of only the digital filter 41A among the digital filters 41A to 41M. However, the other digital filters 41B to 41M have the same structure and perform the same operation. It is.

  The digital filter 41A includes a delay buffer 42A having a plurality of stages of outputs. The amount of delay at each stage of the delay buffer 42A is set according to the arrangement of the microphones 2 in the microphone array and the area in front of the microphone array (area for detecting a speaker). In this example, the delay buffer 42A has four-stage outputs, and these output signals are input to the FIR filters 431A to 434A.

  The delay buffer 42A buffers audio signals obtained by adding different delay times to the audio signal output from the microphone 2A, and outputs the delayed audio signals to the FIR filters 431A to 434A. Here, the delayed audio signals output to the FIR filters 431A to 434A correspond to the respective areas in front of the microphone array. FIG. 3 is a diagram illustrating an example of a sound source direction detection method. FIG. 6A is a diagram showing the relationship between the positional relationship between the sound source and the microphone and the delay when the sound generated from the sound source is picked up by each microphone, and FIGS. FIG. 4 is a diagram showing a concept of forming a delay correction amount based on a delay of a collected audio signal.

  As shown in the figure, in this sound emission and collection device, four partial areas 101 to 104 are set in front of the microphone array. The sound generated in the partial area 101 is first picked up by the nearest microphone 2A. Then, sound is collected by each microphone in order according to the distance between the partial area 101 and the microphone 2, and finally collected by the farthest microphone (microphone 2L in the figure). On the other hand, the sound generated in the partial area 104 is collected first by the nearest microphone 2L, sequentially collected by each microphone according to the distance between the partial area 104 and the microphone 2, and finally collected by the farthest microphone 2A. Sounded. Thus, the sound generated in each region is collected with a delay time (delay) corresponding to the distance from the microphone.

  Here, for the partial area 101, as shown in FIG. 3B, the audio signals collected by the microphones 2A to 2L are subjected to delay processing. That is, the corresponding delay correction amount is set so as to correct the delay shown in FIG. On the other hand, as shown in FIG. 3C, the partial area 104 is subjected to delay processing on the audio signals collected by the microphones 2A to 2L.

  A delayed sound signal for forming a sound collecting beam corresponding to the partial region 101 is generated in the delay buffer 42A and output to the FIR filter 431A. In addition, a delayed sound signal for forming a sound collecting beam corresponding to the partial region 102 is output to the FIR filter 432A. Similarly, a delayed sound signal for forming a sound collecting beam corresponding to the partial area 103 is output to the FIR filter 433A, and a delayed sound signal for forming a sound collecting beam corresponding to the partial area 104 is supplied to the FIR filter 434A. Is output. The delay amounts of these delayed audio signals are set according to the distance between the microphone 2 and each area as shown in FIG. For example, the delayed audio signal corresponding to the partial area 101 has a large delay amount because the distance between the microphone 2A and the partial area 101 is short, and the delayed audio signal corresponding to the partial area 104 has a distance between the microphone 2A and the partial area 104. The delay is small because it is farthest away.

  In FIG. 2, the FIR filters 431A to 434A all have the same configuration, and filter and output the delayed audio signals input thereto. The FIR filters 431A to 434A can set a detailed delay time that cannot be realized by the delay buffer 42A. That is, by setting the sampling period and the number of taps of the FIR filter to desired values, for example, when the sampling period in the delay buffer 42A is an integer part of the delay time, the decimal part of the delay time can be realized. it can.

  The delayed audio signals output from the FIR filters 431A to 434A are amplified by the respective amplifiers 441A to 444A and input to the adders 45A to 45D. The other digital filters 41B to 41M have the same configuration as that of the digital filter 41A, and output delayed audio signals to the adders 45A to 45D in accordance with delay conditions set in advance.

  The adder 45A synthesizes the delayed audio signals input from the digital filters 41A to 41M, and generates a sound collection beam corresponding to the partial area 101 in FIG. Similarly, the adder 45B synthesizes the delayed speech signals input from the digital filters 41A to 41M to generate a sound collection beam corresponding to the partial region 102 in FIG. 3, and the adder 45C The delayed sound signals input from 41A to 41M are synthesized to generate a sound collection beam corresponding to the partial region 103 in FIG. Further, the adder 45D synthesizes the delayed audio signals input from the digital filters 41A to 41M, and generates a sound collection beam corresponding to the partial region 104 in FIG.

  The collected sound beams output from the adders 45 </ b> A to 45 </ b> D are output to a band pass filter (BPF) 46. The BPF 46 filters each sound collection beam and outputs a sound collection beam in a predetermined frequency band to the level determination unit 47. Here, the BPF 46 uses the fact that the frequency band to be beamed differs depending on the width of the microphone array and the installation interval of the microphones 2, and sets the frequency band corresponding to the sound to be collected by each sound collecting beam as the pass band. Set. For example, if the voice to be collected is the voice of the speaker, a frequency band corresponding to the human voice band may be set as the pass band.

  The level determination unit 47 outputs information indicating the level of each sound collecting beam to the controller 8. The controller 8 compares the levels of the input sound collecting beams and selects the sound collecting beam having the highest level. A high level of the sound collecting beam means that a sound source (speaker) exists in a region corresponding to the sound collecting beam, and the region where the speaker is present when divided into the four regions shown in FIG. Can be detected.

  Here, the controller 8 generates information (hereinafter referred to as “speaker position information”) indicating the presence area of the speaker based on the area corresponding to the sound collecting beam having the highest level. When the level (absolute level) of the highest-level sound pickup beam is less than a predetermined threshold (for example, the level of general uttered speech), the controller 8 determines that there is no speaker and sets the speaker position information. It may not be generated.

  Based on the generated speaker position information, the controller 8 selects the sound collection beam corresponding to the speaker position information in the signal selection unit 48 and outputs it to the speech speed conversion unit 5 as a speaker voice signal. Set to. In addition, the controller 8 sets the signal selection unit 48 to select any one of the collected sound beams corresponding to directions other than the area indicated by the speaker position information and output the selected sound beam to the mixer 6 as a background audio signal. To do. The controller 8 may be set so that the signal selection unit 48 selects a plurality of sound collecting beams corresponding to directions other than the region indicated by the speaker position information, combines them, and outputs them to the mixer 6. . Of course, all the collected sound beams corresponding to directions other than the region indicated by the speaker position information may be combined and output to the mixer 6.

Here, depending on the level of each sound collecting beam, the following two patterns can be considered for the speaker voice signal and the background voice signal to be output.
(1) When the background sound is a point sound source In this case, one of the collected sound beams corresponding to directions other than the area indicated by the speaker position information includes a high level. Therefore, as a result of comparing the levels of the respective sound collecting beams, the controller 8 determines that one of the sound collecting beams corresponding to directions other than the area indicated by the speaker position information has a level equal to or higher than a predetermined value (however, the predetermined value). Is detected in the signal selection unit 48 so that the collected sound beam in this direction is output as a background audio signal.
(2) When the background sound is non-localized In this case, a high level is shown for a plurality of sound collecting beams corresponding to directions other than the region indicated by the speaker position information. Accordingly, as a result of comparing the levels of the respective sound collecting beams, the controller 8 is a sound collecting beam corresponding to a direction other than the region indicated by the speaker position information, and a level (for example, a majority) or a level equal to or greater than a predetermined value. However, when a signal less than the predetermined threshold is detected, the signal selection unit 48 is set so that the highest level of these collected beams is output as a background audio signal. At this time, since the background sound component is also included in the sound collection beam corresponding to the speaker position information, the controller 8 determines whether the sound collection beam corresponding to the speaker position information and the adjacent sound collection beam are The signal selection unit 48 is set to output the difference as a speaker voice signal.

  As described above, the audio signal processing unit 4 can separate the speaker's voice and other voices and output them to the subsequent stage.

  2 shows an example in which four partial areas 101 to 104 are set in front of the microphone array and a sound collecting beam is formed for each area. However, the number of output stages of the delay buffer 42 shown in FIG. 2 is increased. By setting the number of FIR filters, amplifiers, and adders as many as the number of output stages of the delay buffer 42, it is possible to form sound collecting beams for a larger number of regions. Also, the microphone arrays are arranged in two rows, and the sound signal processing unit shown in FIG. 2 is connected to each row, so that a sound collecting beam is formed in the front direction of each microphone array. It is also possible to form a sound collecting beam in a direction (that is, a direction of approximately 360 degrees).

  In addition, the controller 8 may extract a voice feature amount from each collected sound beam, and distinguish between a speech voice and a musical sound voice (including a singing voice, for example). The speech feature amount typically represents a speaker formant, pitch, and the like, and is extracted from a frequency spectrum (power spectrum) obtained by Fourier transforming speech data, and a cepstrum obtained by logarithmically transforming the power spectrum and then performing inverse Fourier transform. The voice feature quantity of the utterance voice and the voice feature quantity of the musical sound voice are recorded in the storage unit 3 in advance, and if the voice feature quantity of each sound collection beam matches the voice feature quantity of the utterance voice, this is recorded. If it is selected as an audio signal and matches the audio feature amount of the musical sound, this may be selected as the background audio signal. In addition, when there are a plurality of high-level sound collecting beams, the sound feature amount of each sound collecting beam is analyzed, and the sound feature amount of the speech sound may be determined as the sound collecting beam of the speaker.

  Prior to the conference, the chairperson or the like operates the sound emission and collection device and asks each conference participant to speak to generate speaker location information in advance and record it in the storage unit 3. Also good. In this case, during the conference, the controller 8 selects the sound collection beam corresponding to the speaker position information in the signal selection unit 48 based on the speaker position information stored in the storage unit 3 and selects this. It is set to output to the speech speed conversion unit 5 as a speaker voice signal. Further, the controller 8 selects any one of the sound collecting beams corresponding to directions other than the area indicated by the speaker position information stored in the storage unit 3 in the signal selection unit 48 and uses this as a background audio signal. Set to output to mixer 6.

  Next, the speech speed conversion unit 5 performs a speech speed conversion process on the input speaker voice signal in accordance with an instruction from the controller 8. The speech speed conversion process is not performed simply at low speed, but is performed as follows. That is, in the speech speed conversion process, the voice signal is cut into a waveform of one cycle, a new periodic waveform is generated by combining one section before and after each periodic waveform, and the newly synthesized periodic waveform is generated between each periodic waveform. Insertion is a process of increasing the number of periodic waveforms of a signal and extending the signal along the time axis while maintaining the pitch.

  FIG. 4A is a flowchart showing the procedure of decompression processing. FIG. 2B is a diagram for explaining the expansion method. In FIG. 9A, first, the number of samples in one cycle (sampling frequency × 1 / signal frequency) of the head portion of the input audio signal is detected (S91). Two periodic waveforms, which are sample data for one period, are taken out and, as shown in FIG. 5B, an attenuation wave is created by multiplying the first periodic waveform A by an attenuation gain coefficient, An increasing wave is created by multiplying the second periodic waveform B by an increasing gain coefficient (S92). Then, by adding and synthesizing these, a periodic waveform having an intermediate shape between A and B is synthesized (S93). The synthesized waveform is inserted between the periodic waveform A and the periodic waveform B and output (S94) as shown in FIG.

  When compressing the audio data, as shown in FIG. 5B, by outputting the synthesized waveform of the intermediate shape of A and B synthesized in S93 in place of the periodic waveforms A and B, Audio data can be compressed 1/2 times in the time axis direction.

  Also, the conversion speed can be made variable by defining the cycle for performing the speech speed conversion processing. For example, as shown in FIG. 5C, by synthesizing two periodic waveforms for each period and inserting them between each periodic waveform, the voice data can be expanded twice in the time axis direction, As shown in FIG. 4D, by synthesizing two periodic waveforms every two periods, it can be expanded to 3/2 times.

  In the speech speed conversion, only the head portion (for example, 700 msec) of the voice section is expanded and the subsequent portion is output at the normal speed, so that it is not expanded more than necessary. The head portion may be expanded and the subsequent portion may be compressed. The distinction between the voice section and the noise section may be determined from the periodicity of the voice signal. For example, the correlation value is calculated by dividing the audio signal into a predetermined length and multiplying or subtracting corresponding sample data. As shown in FIG. 6, when this correlation value is lower than a predetermined threshold, it is determined as a noise interval, and when it is higher, it is determined as a speech interval. In the case of an audio signal having a high periodicity such as speech, the correlation value is high, and in the case of an audio signal having a low periodicity such as noise, the correlation value is low.

  In the present embodiment, the speech speed is converted for the first 700 msec of the speech section. However, the speech speed may be converted for a longer section length, or the speech speed may be converted for a shorter section length. May be. Further, the expansion rate may be changed during the section where the speech speed conversion is performed. For example, when the section length is 700 msec, the speech speed may be converted at an expansion rate such that the first 600 msec is expanded by 2 times and the subsequent 100 msec is expanded by 3/2 times.

  The speaker voice signal subjected to the speech speed conversion by the speech speed conversion unit 5 as described above is input to the mixer 6 and mixed with the background speech signal input from the audio signal processing unit 4 in the mixer 6. The mixed audio signal is input to the recording / playback unit 7. The recording / playback unit 7 supplies the input audio signal to the speaker 1 and the input / output I / F 9, converts the audio signal into audio data (for example, compressed data such as MP3), and inputs the audio data to the storage unit 3. . The recording / reproducing unit 7 reads out audio data recorded in the storage unit 3 and supplies an audio signal based on the audio data to the speaker 1 and the input / output I / F 9.

  The speaker 1 emits the audio signal input from the recording / playback unit 7. As the speaker 1, a cone type speaker is generally used, but other types such as a horn type speaker may be used. In FIG. 1, a D / A converter that converts a digital audio signal into an analog audio signal, an amplifier that amplifies the signal, and the like are omitted.

  The storage unit 3 records the audio data input from the recording / playback unit 7. In addition, as described above, the speaker position information input from the controller 8 is also recorded.

  As a result, among the sounds collected by the sound emission and collection device, only the voice of the speaker is converted in the speech speed, and the background sound is emitted or recorded without being converted in the speech speed.

  The input / output I / F 9 supplies an audio signal to another device. The input / output I / F 9 includes an interface according to a device to which the supply is made. For example, the input / output I / F 9 converts an audio signal into information suitable for network transmission, and other sound emission and collection sounds connected via the network interface. An audio signal is output to the device. Further, the input / output I / F 9 inputs an audio signal from another sound emitting and collecting apparatus connected via the network, and inputs this to the recording / reproducing unit 7. The recording / playback unit 7 records the sound collected by the own device and the sound input from another device in the storage unit 3.

  In the above embodiment, the single speaker 1 is shown as the sound emitting side. However, a plurality of speakers 1 may be arranged in a straight line to constitute a speaker array. In this case, by sequentially delaying the audio signal supplied to each speaker, the audio beam can be focused, and sound image localization as if the audio was emitted from the position of the speaker can be achieved.

  In addition, when the collected audio signal is output to another device and the speaker array is configured on the other device side, the above-described speaker position information is also output, so that the voice is also transmitted to the position of the speaker in the other device. Sound image localization as if it were emitted from

  Further, when a plurality of sound emitting and collecting apparatuses of the above embodiment are connected via a network, the following application examples are possible. FIG. 7 is a diagram showing an example in which an audio conference system is configured by connecting a plurality of sound emitting and collecting apparatuses of the above-described embodiment via a network. This audio conference system includes sound emission and collection devices 111A to 111C connected via a network 100. Since the sound emission and collection devices 111A to 111C have the same configuration and function as the sound emission and collection device described in the above embodiment, detailed description of each configuration and function is omitted.

  The sound emission and collection devices 111 </ b> A to 111 </ b> C are disposed at points a to c that are separated from each other. A sound emitting and collecting device 111A is disposed at the point a, a sound emitting and collecting device 111B is disposed at the point b, and a sound emitting and collecting device 111C is disposed at the point c.

  At the point a, the participants A and B are present in the directions Dir11 and Dir13 with respect to the sound emitting and collecting apparatus 111A, respectively. At the point b, the sound source A is present in the direction Dir22 with respect to the sound emission and collection device 111B. At the point c, the conference persons C and D are present in the directions Dir31 and Dir32 with respect to the sound emission and collection device 111C, respectively. The azimuth Dir11 to Dir14, the azimuth Dir21 to Dir24, and the azimuth Dir31 to Dir34 respectively correspond to the four partial areas 101 to 104 in the above-described embodiment, and the sound emitting and collecting apparatus collects sound in these directions. To do.

  In this audio conference system, each sound emitting and collecting device transmits the sound collected by its own device to all other sound emitting and collecting devices. In addition, each sound emission and collection device records the sound transmitted from the other device together with the sound collected by its own device.

  When the conference participant A and the conference participant B speak, the sound emission and collection device 111A converts the speed of these voices and then transmits the speech to another device. In addition, when the conference person C and the conference person D speak, the sound emission and collection device 111C converts these voices and then transmits them to other devices.

  Here, the sound emission and collection device 111B outputs the musical sound generated by the sound source A to another device without converting the speech speed. At this time, the sound emission and collection device 111B transmits the sound without converting the speech speed even when the level of the musical sound generated by the sound source A is very high. For example, the speech speed is not converted even if the level exceeds a predetermined threshold value (general speech level). That is, in FIG. 1, the controller 8 always outputs the collected sound to the mixer 6 to the audio signal processing unit 4 when given an instruction not to convert the speech speed from an operation unit (not shown) or the like. Set. As a result, in this sound emission and collection device, sound that does not always convert the speech speed is output. In this case, since the controller 8 outputs the sound collecting beam having the highest level, the controller 8 does not need to determine the absolute value of the sound collecting beam level (whether it is equal to or higher than the level of a general uttered voice).

  Note that the controller 8 may set the audio signal processing unit 4 to always output the collected sound to the speech speed conversion unit 5. In this case, in this sound emitting and collecting apparatus, the speech speed converted voice is always output.

  In this way, even if an arbitrary sound emitting and collecting device in the audio conference system is used as a device dedicated to background sound output (a sound emitting and collecting device that does not convert the speech speed), While listening at normal speed, you can listen to the speaker's voice slowly. In each audio conference device, the background sound is recorded at the normal speed, and only the voice of the speaker is converted and recorded.

The block diagram which shows the structure of the sound emission and collection apparatus of embodiment of this invention. Block diagram showing the configuration of the main part of the audio signal processing unit Diagram showing sound source detection area Diagram showing speech speed conversion processing The figure which shows the speech speed conversion processing when changing the expansion rate Diagram showing an example of calculating the correlation value of input audio data The figure shown about the example which connects multiple sound emitting and collecting apparatuses of the said embodiment via a network, and comprises an audio conference system.

Explanation of symbols

1-speaker 2-microphone 3-storage unit 4-audio signal processing unit 5-speech speed conversion unit 6-mixer 7-recording / reproducing unit 8-controller

Claims (4)

A microphone array in which a plurality of microphones are arranged;
A sound collection control unit that forms a sound collection beam with respect to a plurality of user directions and identifies a speaker orientation by comparing the sound collection beam intensities;
A sound signal selection means for selecting the sound beam collected in the speaker direction as a speech sound signal, and a sound beam other than the sound beam collected in the speaker direction as a background sound signal;
Speech speed converting means for converting the speech speed of the speech signal;
A mixer that mixes the utterance voice signal converted by the speech speed conversion means and the background voice signal selected by the voice signal selection means;
A sound collecting device.   The sound signal selecting means, when there is a sound collecting beam of a predetermined level or more in a direction other than the sound collecting beam selected as the speech sound signal, selects only the sound collecting beam in that direction as a background sound signal. The sound collecting device according to 1.   The voice signal selecting means uses the difference signal between the sound collection beam selected as the speech voice signal and the sound collection beam in the direction adjacent to the sound collection beam selected as the speech voice signal as the speech signal. The sound collecting device according to claim 1, wherein the sound collecting device is input to the speech speed converting means. Further comprising speech signal extraction means for extracting speech signals of speech from a plurality of sound collection beams formed by the sound collection control unit;
2. The sound collection control unit determines a direction of a sound collection beam having a highest level among a plurality of sound collection beams and a direction of the sound collection beam from which the speech signal extraction unit has extracted a speech signal as a speaker orientation. The sound collecting device according to claim 2 or claim 3.

Images