JP2008292621A - Speech speed conversion device, speaking device and speech speed conversion method - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a speech speed conversion device, a speaking device and a speech speed conversion method, capable of automatically converting speech speed of only speech of a speaker specified by a student, with a conversion rate by which a user feels most desirable. <P>SOLUTION: The speech speed conversion device is provided with: a speech feature extraction section 101 for extracting a speech feature of the individual speaker who attends speaking; a speech feature storage section 102 for storing the speech feature for the speaker specified by the student; a speaker determination section 103 for determining whether or not an uttering person is one of the speaker specified by the student by comparing the speech feature of a current speaker with the speech feature stored by the speech feature storage section 102; and a speech speed conversion section 104 for converting the speech speed of received speech of the speaker, when it is determined that the uttering person is the speaker specified by the student by the speaker determination section 103. <P>COPYRIGHT: (C)2009,JPO&INPIT

Description

  The present invention relates to a speech speed conversion device and a speech speed conversion method that can be used for a telephone device that connects a plurality of bases and performs a voice call such as a conference phone. The present invention also relates to a call device having a speech speed conversion device.

  The technology of speech speed conversion that changes the speed of the voice without changing the pitch of the voice is conventionally known, and is widely used for voice reproduction apparatuses such as IC (Integrated Circuit) recorders, telephones, televisions, radios, and the like. It's being used. The content of the speech speed conversion process is shown in, for example, (Patent Document 1).

  A general processing configuration for converting speech into a slow speech speed will be described. FIG. 18 is a block diagram schematically showing a configuration of a conventional speech speed conversion device. In FIG. 18, 201 is an audio pitch detector for detecting the pitch (fundamental frequency) of an audio signal, 202 is an insertion waveform generator for extracting an audio waveform in units of the audio pitch period, and 203 is an input audio signal for the extracted pitch waveform. It is a waveform connection part that expands the waveform in the direction of the time axis by inserting it periodically. The voiced signal waveform composed of the fundamental frequency and its overtones tends to repeat the same waveform periodically in pitch units, so this configuration should be converted to a slower speech speed without changing the pitch. Is possible.

  Consider a case where speech rate conversion processing is applied to a device that performs real-time (no delay allowed) voice transmission / reception, such as a communication device. For example, if the speech speed of the received voice is converted to a slower rate and played continuously, the delay between the time when the received voice is actually generated and the playback time after the conversion of the speech speed increases over time. It will cause a significant malfunction. Therefore, when applied to such a device, special devices have been conventionally made. The concept will be described with reference to FIG. In real-time speech speed conversion processing, conversion is performed to slow down the speech interval (interval where the speech signal is present in the received signal), and compression is performed during the silent interval (interval where the speech signal is not present in the received signal). Therefore, the delay from the actual speech time is reduced.

Further, when connecting signals, smoothing is performed by providing overlapping sections little by little to prevent the occurrence of distortion. When a device that performs such a speech rate conversion process is installed in a telephone conference device, it is possible to convert the voice of a speaker that is difficult to hear quickly into a speech rate that is slow and easy to hear. In addition, when a foreign language conversation is conducted at an international conference or the like, the intelligibility of the foreign language is improved, and the effect of contributing to the smooth progress of the conference can be expected.
JP-A-5-257490

  However, according to the above-described conventional technology, when the speech speed conversion process is applied to a telephone device used for a teleconference in which a plurality of speakers participate, the speaking speed of all the speakers is converted slowly. . Generally speaking, there is an individual difference in speaking speed, and a fast-speaking speaker improves the intelligibility of the conversation by converting the speaking speed to a slower speed. However, the speaking speed of the speaker who originally has a slow speaking speed and does not need to convert the speaking speed. Since the voice is also converted together, the intelligibility of the conversation may be impaired. In addition, there is a desire to selectively convert a speaker's voice to a slower speaking speed, such as when a specific speaker among multiple speakers speaks a foreign language, but this is not possible with the conventional configuration. there were. As described above, when the conventional speech speed conversion process is applied to a telephone conference in which a plurality of speakers participate, there is a problem that a good effect cannot always be obtained.

  As described above, in the speech rate conversion device and the speech rate conversion method installed in the communication device in which multiple speakers can participate, the speech of a specific speaker, such as a speaker with a high speech rate or a speaker speaking a foreign language, can be obtained. It is required to perform speech rate conversion at an optimal speech rate conversion rate.

  The present invention has been made in view of the above, and is capable of automatically converting the speech speed of only the voice of the speaker specified by the listener at a conversion rate that the user feels most desirable. An object is to provide a device, a communication device, and a speech speed conversion method.

  In order to solve the above-mentioned problems, the present invention provides a voice feature extracting means for extracting voice features of individual speakers participating in a call, and a voice feature for storing voice features extracted for a speaker designated by a listener. Whether the speaker is one of the speakers designated by the listener by comparing the voice features of the current speaker with the voice features stored in the voice feature storage means And a speech speed conversion means for converting the speech speed of the speech received by the speaker when the speaker determination means determines that the speaker is designated by the listener. , With.

  The present invention also extracts the voice features of individual speakers participating in the call, stores the voice features extracted for the speaker specified by the listener, and stores the voice features of the current speaker. It is determined whether or not the speaker is one of the speakers specified by the listener by comparing with the voice feature. As a result of this determination, when it is determined that the speaker is designated by the listener, speech speed conversion processing for converting the speech speed of the received voice of the speaker is performed.

  According to the present invention, only the voice of the speaker designated by the listener is automatically converted at the conversion rate that the user feels most comfortable. Therefore, even in a call in which a plurality of speakers participate, Thus, it is possible to obtain a speech speed conversion device that can improve the intelligibility of the utterance content of the speaker by the listener. In addition, since the speech speed is not converted for the voice of a speaker not specified by the listener, the speech speed can improve the understanding level of the speaker's speech overall for the listener. A conversion device is obtained.

  Also, according to the present invention, only the voice of the speaker specified by the listener is automatically converted at the conversion rate that the user feels most comfortable, so the conversation in which a plurality of speakers participate. The speech speed conversion method that can improve the intelligibility of the content of the speaker's utterance by the listener is also obtained. In addition, since the speech speed is not converted for the voice of a speaker not specified by the listener, the speech speed can improve the understanding level of the speaker's speech overall for the listener. A conversion method is obtained.

  According to a first aspect of the present invention, there is provided a speech speed conversion device for extracting speech features of individual speakers participating in a call, and a speech feature for storing speech features extracted for a speaker designated by a listener. Talk to determine whether the speaker is one of the speakers specified by the listener by comparing the voice features of the storage means and the current speaker with the voice features stored in the voice feature storage means And a speech speed conversion means for converting the speech speed of the received voice of the speaker when the speaker determination means determines that the speaker is the speaker designated by the listener. Even when there are a plurality of talkers, only the voice of the talker designated by the listener is converted to a good talk speed.

  According to a second aspect of the present invention, there is provided a speech rate conversion device according to the first aspect, wherein the designated speaker conversion condition storage means stores the conversion rate of the optimal speech rate set for each speaker specified by the listener. And a designated speaker conversion condition selecting means for selecting the conversion rate of the speech speed corresponding to the speaker determined by the speaker determining means from the designated speaker conversion condition storage means, and the speech speed converting means is , Converting the speaking speed of the speaker specified by the listener using the conversion rate of the speaking speed selected by the designated speaker conversion condition selecting means, and for each speaker of the other party specified by the listener It has the effect that the conversion rate of speech speed can be set.

  According to a third aspect of the present invention, in the second aspect, the designated speaker conversion condition storage means further stores the amplification factor of the reproduction volume set for each of the speakers designated by the listener. The designated speaker conversion condition selecting means selects the conversion rate of the speech speed and the amplification factor of the reproduction volume corresponding to the speaker determined by the speaker determining means from the specified speaker conversion condition storage means, and The speed conversion means converts the speech speed and reproduction volume of the speaker designated by the listener using the conversion rate of the speech speed selected by the designated speaker conversion condition selection means and the amplification factor of the reproduction volume. Thus, it has an effect that the amplification factor of the reproduction volume can be set together with the conversion rate of the speech speed for each of the other party's speakers designated by the listener.

  The communication device of the fourth invention of the present invention is connected to another communication device via a communication line and communicates, a communication device for collecting communication, a sound collection device for collecting the voice uttered by the speaker, There is provided a voice output means for reproducing and outputting a voice from a call device and a speech speed conversion device according to any one of the first to third inventions, and there are a plurality of talkers. Even in such a case, only the voice of the speaker designated by the listener is converted into a good speech speed and / or reproduction volume.

  Furthermore, the speech speed conversion method according to the fifth aspect of the present invention performs speech extraction processing for extracting the speech features of individual speakers participating in the call, and the speech features extracted for the speakers designated by the listener. To determine whether or not the speaker is one of the speakers specified by the listener by comparing the voice features of the current speaker with the stored voice features. Performs speaker determination processing, and performs speech speed conversion processing to convert the speech speed of the received speech of the speaker when it is determined that the speaker is designated by the listener as a result of the determination. Even when there are a plurality of speakers, only the voice of the speaker designated by the listener is converted to a good speech speed.

  According to a sixth aspect of the present invention, there is provided a speech speed conversion method for performing a designated speaker conversion condition storing process for storing a conversion rate of an optimal speech speed set for each of speakers specified by a listener, and determining a speech speed. The designated speaker conversion condition selection process is performed to select the conversion rate of the speech speed corresponding to the speaker determined by the speaker determination process before the process from the contents stored in the designated speaker conversion condition storage process. In the conversion process, the speaker's speech speed specified by the listener is converted using the conversion rate of the speech speed selected in the designated speaker conversion condition selection process. It has the effect that the conversion rate of the speech speed can be set for each speaker.

  According to a seventh aspect of the present invention, in the designated speaker conversion condition storing process, the reproduction volume gain set for each of the speakers designated by the listener is further stored, and the designated speaker conversion is stored. In the condition selection process, the conversion rate of the speech speed corresponding to the speaker determined by the speaker determination process and the amplification factor of the reproduction volume are selected from the contents stored in the designated speaker conversion condition storage process, and the speech speed conversion is performed. In the process, the speaker speed specified by the listener and the playback volume are converted using the conversion rate of the speech speed selected in the specified speaker conversion condition selection process and the amplification factor of the playback volume. In addition, it has an effect that the amplification factor of the reproduction volume can be set together with the conversion rate of the speech speed for each speaker of the other party specified by the listener.

  Embodiments of the present invention will be described below.

(Embodiment 1)
In the first embodiment, in a conference call in which a plurality of speakers participate, it is determined who is the speaker, and the determination result is compared with the speech rate conversion rate set for each speaker. A speech speed conversion device, a speech device, and a speech speed conversion method for determining a speed conversion rate, and a speech speed conversion unit automatically converting a speech speed of an individual speaker into a speech speed preferable for a listener according to the result explain.

  FIG. 1 is a perspective view showing an example of a communication device including the speech speed conversion device according to Embodiment 1 of the present invention, and FIG. 2 is a top view of the communication device of FIG. 1 to 2, reference numeral 601 denotes a communication device, reference numerals 602a to 602d denote microphones that collect the user's voice, reference numeral 603 denotes a speaker that reproduces the received voice, reference numeral 604 denotes a communication cable connected to the partner line, and reference numeral 605 denotes a communication cable. A registration button for the user (corresponding to the listener in the claims) to specify a speaker to which the speech speed conversion is applied, 606 is a slow playback button for starting / ending the speech speed conversion, and 607 is an outgoing call An operation button for performing an incoming call operation, and a display unit 608 displays a call state. Here, the speaker corresponds to the sound output means in the claims, and the microphones 602a to 602d also correspond to the sound collecting means.

  FIG. 3 is a block diagram schematically showing the configuration of components related to the speech speed conversion apparatus according to the first embodiment of the present invention in the communication apparatus of FIGS. In FIG. 3, reference numeral 701 denotes a digital signal processor (Digital Signal Processor, hereinafter referred to as DSP) for performing various operations and control of peripheral devices, 702 a communication path interface, 703 an operation interface, 704 a memory, and 705 a digital / analog. (Hereinafter referred to as D / A) converter, 706 is an analog / digital (hereinafter referred to as A / D) converter, 707 is a speaker, 708 is a microphone, 7011 is a received signal input unit, 7012 is an audio signal output unit, and 7013 is communication. A path 7014 is a reception signal path.

  An outline of the operation of the call device in FIG. 3 will be described. The communication path interface 702 of the call device 601 is connected to one end of the communication path 7013, and transmits and receives a call signal (voice signal) with another call device (not shown) connected to the other end of the communication path 7013. A voice signal transmitted from another call device connected to the other end is input from the reception signal input unit 7011 to the DSP 701 from the communication path 7013 via the communication path interface 702 and the reception signal path 7014. In the DSP 701, predetermined processing such as speech speed conversion according to the first embodiment is performed on the received audio signal.

  Here, for example, when the signal transmitted / received in the communication path 7013 is a digitized signal such as an IP (Internet Protocol) packet, the signal processed in the communication path interface 702 is, for example, PCM (Pulse-Code Modulation). The reception signal path 7014 is usually a serial bus or a parallel bus. When the communication path 7013 is an analog signal path such as PSTN (Public Switched Telephone Networks), an A / D converter 706 is required in the communication path interface 702 or the DSP 701. When the A / D converter 706 is included in the communication path interface 702, all subsequent processing is performed using digital signals as before, and the reception signal path 7014 is usually a serial bus or a parallel bus. On the other hand, when the A / D converter 706 is built in the DSP 701, the reception signal path 7014 is an analog signal line.

  The received voice whose speech speed has been converted by the DSP 701 is converted into an analog signal by the D / A converter 705 and reproduced from the speaker 707. Similarly, the user's transmitted voice is collected by the microphone 708, converted into a digital signal by the A / D converter 706, passed to the DSP 701, and transmitted to the communication path interface 702 as a transmitted signal (voice signal).

  FIG. 4 is a block diagram schematically showing a functional configuration of the speech rate conversion apparatus according to Embodiment 1 of the present invention. In FIG. 4, 101 is a voice feature extraction unit that extracts the voice features of individual speakers participating in the call, 102 is a voice feature storage unit that stores the extracted voice features, and 103 is the voice of the current speaker. A speaker determination unit that determines whether the speaker is one of the speakers specified by the listener by comparing with the voice feature stored in the voice feature storage unit 102, and 104 is a speaker determination unit When it is determined that the speaker is designated by the listener in 103, the speech speed converting unit converts the speech speed of the received voice of the speaker into a predetermined speed.

  In the first embodiment, the speech feature storage unit 102 corresponds to the memory 704, and the speech feature extraction unit 101, the speaker determination unit 103, and the speech speed conversion unit 104 are stored in the memory 704 in FIG. It corresponds to the software program that runs above. That is, the speaker determination unit 103 and the speech speed conversion unit 104 calculate the received voice received by the DSP 701 from the communication path interface 702 according to the program stored in the memory 704 and output the result to the D / A converter 705. Is realized. The voice feature extraction unit 101 is realized by calculating a voice feature amount from a received voice received by the DSP 701 from the communication path interface 702 according to a program stored in the memory 704, and a voice feature storage unit. 102 is realized by storing the voice feature amount acquired by the voice feature extraction unit 101 in the memory 704.

  A description will be given of a reception voice reproduction process in the communication device of the first embodiment configured as described above. FIG. 5 is a flowchart showing an example of the procedure of the speech speed conversion method according to Embodiment 1 of the present invention. Here, as a premise, it is assumed that the call device is connected to another point where a plurality of speakers exist and is talking (for example, during a conference call).

  First, when it is necessary for the user of the communication device 601 to specify the voice of the other party's speaker that he / she wants to hear by converting the speech speed (Yes in Step 1), such a speaker's While speaking, the registration button 605 is pressed to instruct the call device 601 to register voice features. Thereby, the voice feature extraction unit 101 calculates the feature amount of the received voice at that time and stores it in the voice feature storage unit 102 (step 2).

  The speech feature amount extracted in step 2 is assumed to be speech spectrum, pitch frequency, speech formant transition information, etc., which are generally used for speaker authentication. For example, the spectrum is extracted by FFT (Fast Fourier Transform) calculation of the audio signal, the pitch frequency is extracted by the cepstrum method or the correlation function method, and the time transition of the audio formant is extracted by calculating the outline of the spectrum. Then, it is converted into a format suitable for the matching method of the speaker determination process. For example, the DP (Dynamic Programming) matching method for calculating the spectral distance of word utterance is converted as time-series data of spectrum information, and the matching method based on a probability model such as HMM (Hidden Markov Model) is converted as state transition probability information. .

  After that, or in the case where the speaker who is speaking in step 1 is not registered as the other party who wants to hear by converting the speech speed (No in step 1), the user performs another registration of the speaker whose speech speed is to be converted. In the case (No in step 3), the process returns to step 1 and the above-described processing is repeated until all the speakers whose speech speed is to be converted are registered.

  The registration of the speaker who wants to change the speech speed by the user is completed (Yes in Step 3), and it is determined by the communication device 601 whether or not the slow playback button 606 has been pressed by the user during normal conversation. (Step 4). When the slow playback button 606 is pressed (Yes in Step 4), the voice feature extraction unit 101 of the call device 601 continues to calculate the voice feature amount of the received voice (Step 5).

  After that, the speaker determination unit 103 calculates the distance between the voice feature amount of the received voice calculated in step 5 and the voice feature amount stored in the voice feature storage unit 102, and the result is registered in step 2. It is determined whether or not the speaker is one of the speakers (step 6). That is, if the distance from the speech feature amount is smaller than the threshold (Yes in Step 6), the speaker having the received speech currently being received is determined to be the speaker designated by the user, and the speech speed A flag to that effect is passed to the conversion unit 104. If the distance from the voice feature amount is equal to or greater than the threshold, the speaker determination unit 103 determines that the speaker is not designated by the speaker user who has the received voice currently being received, and the speech speed conversion unit 104 Pass a flag to that effect.

  The speech speed conversion unit 104 performs processing on the received voice based on the flag received from the speaker determination unit 103. That is, when the voice characteristics of the registered speaker are close (Yes in Step 6), the speech speed conversion unit 104 converts the received voice into a speech speed that is easy to hear and reproduces it (Step 7). On the other hand, when it is not close to the voice feature of the registered speaker (No in Step 6) or when the slow playback button 606 is not pressed in Step 4 (No in Step 4), the speech speed conversion unit. 104 performs a predetermined process without converting the speech speed of the received voice. And the process of step 1-step 7 is repeatedly performed until a telephone call is complete | finished.

  In the process shown in FIG. 5 above, the process from step 1 to step 4 for registering a speaker to be speech speed converted and setting whether to perform the speech speed conversion requires an operation process by the user. The selective speech speed conversion processing during a call in steps 5 to 7 is continuously performed automatically by the call device 601 based on the set contents.

  Here, the concept of real-time speech speed conversion in the first embodiment will be described. FIG. 6 is a diagram showing an operation concept of the speech speed conversion process. In FIG. 6, two modes can be selected: a “normal speed” mode in which the slow playback button 606 is not pressed (no speech speed conversion = 100%) and a “slow” mode in which the slow playback button 606 is pressed. Shows the case.

  As shown in FIG. 6, when the “slow” mode is registered so as to have a speech rate conversion rate higher than the normal mode with respect to the “normal speed” mode, the sound period is set for each speaker. The audio signal is expanded in the time direction at the conversion rate. However, if the speech speed of the received voice is uniformly reduced and played continuously, the delay of the playback time after the speech speed conversion with respect to the time when the received voice was actually generated increases with time, which causes a significant problem with the conversation. Will occur. Therefore, by compressing the silent section in accordance with the actual speech timing, the speech speed can be converted without causing a large delay that hinders the conversation even when the conversion rate of the speech speed is different.

  Next, a specific method for realizing the speech speed conversion process will be described with reference to FIGS. FIG. 7 is a diagram showing an example of a configuration when using the communication device according to Embodiment 1 of the present invention, and FIG. 8 is another example of a configuration when using the communication device according to Embodiment 1 of the present invention. FIG. 9 is a diagram illustrating an actual operation example of the speech speed conversion processing in FIGS. 7 to 8. In these examples, a case where two call devices are connected via a communication line and a voice call is performed is illustrated.

  In FIG. 7, 601a is a communication device used by speaker A, 601b is a communication device used by speaker B, 1201a and 1201b are gateways, and 1202 is the Internet. In the configuration example shown in FIG. 7, the two communication devices 601a and 601b are connected to the Internet 1202 via gateways 1201a and 1201b, respectively. In the case of FIG. 7, the audio signal transmitted / received between the call device 601a and the call device 601b is data obtained by packetizing a digital signal.

  Of course, the configuration in use in the first embodiment is not limited to this, and other terminal devices, hubs, routers, and other communication devices may be connected to the gateways 1201a and 1201b, for example. In addition, communication devices such as other terminal devices, hubs, and routers may be connected between the gateway 1201a and the call device 601a or between the gateway 1201b and the call device 601b.

  In FIG. 8, 1301a and 1301b are connection lines connected to the communication devices 601a and 601b, respectively. 1302a and 1302b are modems, 1303a and 1303b are public line networks, 1304a and 1304b are Internet services to which the communication devices 601a and 601b belong. A provider (indicated as ISP in the figure).

  As shown in the configuration example shown in FIG. 8, communication apparatuses 601a and 601b are connected to the Internet 1202 via modems 1302a and 1302b, public line networks 1303a and 1303b, Internet service providers 1304a and 1304b, respectively. It may be. In this case, analog voice signals are transmitted and received on the connection line 1301a between the modem 1302a and the communication device 601a and on the connection line 1301b between the modem 1302b and the communication device 601b. The signal may be digitized and modulated / demodulated. Further, when the connection lines 1301a and 1301b are LAN (Local Area Network) cables and the voice data packets already digitized in the communication apparatuses 601a and 601b are transmitted on the connection lines 1301a and 1301b, the modem 1302a. , 1302b, only modulation / demodulation may be performed.

  7 and 8 are examples, and the configuration shown in FIG. 7 and the configuration shown in FIG. 8 may be mixed, or the modems 1302a and 1302b in FIG. You may have the structure connected only through the same public line.

  Also, the communication devices 601a and 601b in the first embodiment are connected to the microphones (corresponding to the microphones 602a to 602d of the communication device 601 in FIG. 2) incorporated in the communication devices 601a and 601b. The voice of the user 601b is not output to the speakers (corresponding to the speaker 603 of the call device 601 in FIG. 2) built in the call devices 601a and 601b. This is because howling is likely to occur when the voice of the user of the communication device 601a, 601b input to the built-in microphone is output from the speaker of the communication device 601a, 601b. However, if a device that does not generate howling can be configured, the voice of the user of the communication device 601a, 601b input to the built-in microphone is output from the speaker of the communication device 601a, 601b. It may be.

  An operation example of actual speech speed conversion processing in the above configuration will be described with reference to FIG. Here, the speaker A uses the communication device 601a of FIG. 7, and the speaker B uses the communication device 601b. Before starting the operation of FIG. 9, the speaker A presses the registration button (corresponding to the registration button 605 of the call device 601 in FIG. 2) as described above, and steps 1 to 3 in the flowchart of FIG. Then, the voice feature of the speaker B is registered in the call device 601a, and the slow playback button (corresponding to the slow play button 606 of the call device 601 in FIG. 2) is further pressed, and the voice of the speaker B is heard in the call device 601a. It is assumed that the feature amount has been calculated.

  As indicated by reference numerals 1411 and 1413 in FIG. 9, the speed of the voice speaking from the speaker A to the speaker B in the telephone conversation device 601a is the same as the speed of the voice of the speaker A reaching the speaker B from the telephone conversation device 601b. It is. This is because the voice feature of the speaker A is not registered in the talking device 601b on the speaker B side, or the slow playback button 606 of the talking device 601b on the speaker B side is pressed even if it is registered. This is because the voice feature amount is not calculated in the communication device 601b.

  In the former case, since the processing state of the program shown in the flowchart of FIG. 5 installed in the talking device 601b on the speaker B side has not registered voice characteristics at the time of determination, step 1 (No) → step 3 ( Yes) → Transfer to Step 4. Therefore, if the slow playback button 606 has been pressed, the process proceeds in the order of step 4 (Yes) → step 5 → step 6. However, in step 6, the speaker determination unit 103 in FIG. 4 determines “No” based on the signal from the speech feature extraction unit 101. That is, since the voice feature is not registered in the voice feature storage unit 102, the speaker determination unit 103 does not determine that the voice feature is close to the registered speaker's voice feature. Therefore, the speech speed conversion unit 104 in FIG. 4 that has received the determination result does not perform the speech speed conversion of the voice signal transmitted from the communication device 601a. Thereafter, the processing state of the program that executes the flowchart of FIG. 5 mounted on the communication device 601b shifts from Step 6 (No) to Step 1.

  In the latter case, since the processing state of the program shown in the flowchart of FIG. 5 installed in the communication device 601b on the speaker B side is registered at the time of determination, first of all, step 1 (Yes) → The process proceeds from step 2 to step 3 (Yes) to step 4. However, since the slow playback button 606 has not been pressed, “No” is determined in step 4, and the process proceeds to step 1. The state transition shown above is repeatedly performed by the program shown in the flowchart of FIG. 5 installed in the communication device 601b.

  On the other hand, the voices 1412a and 1414a heard by the speaker A from the call device 601a are slower than the voices 1412b and 1414b that the speaker B is talking to the speaker A in the call device 601b. This is because the voice feature of the speaker B is registered in the talking device 601a on the speaker A side, the slow playback button 606 of the talking device 601a on the speaker A side is pressed, and the voice feature amount is obtained in the talking device 601a. This is because the calculation is performed.

  This processing state will be described with reference to the flowchart shown in FIG. When the speaker B is not speaking to the speaker A, that is, in the state shown in the sections 1401 and 1403, the processing state of the program shown in the flowchart of FIG. Is not registered, the process proceeds from step 1 (No) to step 4. Further, since the slow playback button 606 has been pressed, the process proceeds in the order of Step 4 (Yes) → Step 5 → Step 6. In step 6, the speaker determination unit 103 in FIG. 4 determines “No” as a result of comparing the signal from the speech feature extraction unit 101 and the data in the speech feature storage unit 102, so the speech speed conversion unit that receives the determination result 104 does not convert the speech speed of the audio signal sent from the communication device 601b. Thereafter, the processing state of the program shown in the flowchart of FIG. 5 installed in the communication device 601a shifts from step 6 (No) to step 1. The state transition shown above is repeatedly performed by the program shown in the flowchart of FIG. 5 installed in the communication device 601a.

  Next, when talking from the speaker B to the speaker A (when in the state shown by the sections 1402 and 1404), that is, when the voice 1412b or 1414b in FIG. 9 is transmitted from the communication device 601b to the communication device 601a, The processing state of the program shown in the flowchart of FIG. 5 installed in the apparatus 601a shifts in the order of Step 1 (No) → Step 3 (Yes) → Step 4 (Yes) → Step 5 → Step 6 in the same manner as before. Then, as a result of comparing the signal from the speech feature extraction unit 101 and the data in the speech feature storage unit 102 by the speaker determination unit 103 in FIG. 4 in step 6, “speaker B”, that is, “Yes” in step 6. And the state transitions to step 7. Upon receiving the determination result, the speech speed conversion unit 104 in FIG. 4 converts the speech speed of the voice signal 1412b or 1414b of the speaker B sent from the communication device 601b to be 1412a or 1414a, respectively. Thereafter, the processing state of the program shown in the flowchart of FIG. 5 installed in the communication device 601a shifts from step 6 to step 1.

  In the communication device 601a, the user of the communication device 601a can determine in advance the setting on the menu screen or the like as to how much the speech speed conversion speed of the voice signal of the speaker B sent from the communication device 601b is to be set. Alternatively, the device manufacturer may determine a predetermined speech speed conversion speed in advance at the design stage of the communication device 601a.

  In the first embodiment, the case where only the speaker A registers the voice feature of the speaker B and then presses the slow playback button 606 has been described. The slow playback button 606 may be pressed not only by the speaker A but also by the speaker B. Further, in the above-described example, the case where two communication devices 601 are connected is described as an example, but three or more communication devices 601 may be used, or when a plurality of speakers speak in a certain communication device 601. However, it is also possible for another call device 601 to register the voice characteristics of any of the plurality of speakers and to convert the speech speed.

  As described above, in the first embodiment, the speech feature extraction unit 101 that extracts speech features from the speech signal of the received speech, and whether or not the speech features extracted by the speech feature extraction unit 101 belong to the registered speaker. Provided with a speaker determination unit 103 for determining the speech rate, and a speech speed conversion unit 104 that converts the speech speed of the voice signal to a predetermined speech speed when it belongs to a registered speaker. In a call in which a participant participates, only the voice of a speaker designated by the listener, such as a fast-talker or a speaker who speaks a foreign language, is converted to an appropriate speed. As a result, there is an effect that the intelligibility of the utterance contents of those speakers by the listener is improved. On the other hand, since the speech speed conversion is not applied to the voice of the speaker that is not specified by the listener, the conversion becomes counterproductive, that is, the voice of the slow speaker is further delayed, resulting in a decrease in the intelligibility of the conversation. This also has the effect of avoiding this.

(Embodiment 2)
FIG. 10 is a block diagram schematically showing a configuration of a portion related to the speech speed conversion apparatus according to the second embodiment of the present invention. In FIG. 10, as in the first embodiment, 101 is a voice feature extraction unit that extracts voice features of individual speakers participating in a call, 102 is a voice feature storage unit that stores the extracted voice features, 103 Determines whether the speaker is one of the speakers designated by the listener by comparing the voice of the current speaker with the voice features stored in the voice feature storage unit 102. Reference numeral 104 denotes a speech speed conversion unit that receives the determination result of the speaker determination unit 103 and converts the speech speed of the received voice.

  Reference numeral 105 denotes a speaker / speech speed correspondence storage unit that stores the conversion rate of the optimum speech speed set for each speaker. Reference numeral 106 denotes a story corresponding to the speaker determined by the speaker determination unit 103. This is a speech speed selection unit that selects a speed conversion rate from the speaker / speech speed correspondence storage unit 105. Here, the speech feature stored in the speech feature storage unit 102 and the conversion rate of the speech speed stored in the speaker / speech speed correspondence storage unit 105 are attached to, for example, a speaker as a target of the speech feature. Are associated by speaker identification information. The speech speed conversion unit 104 converts the speech speed of the corresponding speaker based on the speech speed conversion rate selected by the speech speed selection unit 106. The designated speaker conversion condition storage means in the claims corresponds to the speaker / speech speed correspondence storage unit 105, and similarly, the designated speaker conversion condition selection means corresponds to the speech speed selection unit 106. Note that the configuration of the communication device including the speech speed conversion device according to the second embodiment is the same as that shown in FIGS. 1 to 3 of the first embodiment, and thus the description thereof is omitted.

  In the second embodiment, the voice feature storage unit 102 and the speaker / speech speed correspondence storage unit 105 correspond to the memory 704 in FIG. The speaker determination unit 103, the speech speed conversion unit 104, and the speech speed selection unit 106 are stored in the memory 704 as software programs that operate on the DSP 701 in FIG. That is, the speaker determination unit 103, the speech rate conversion unit 104, and the speech rate selection unit 106 calculate the received speech received by the DSP 701 from the communication path interface 702 according to a program stored in the memory 704 and perform a D / A converter. This is realized by outputting to 705. The voice feature extraction unit 101 is realized by calculating the voice feature amount from the received voice received by the DSP 701 from the communication path interface 702 according to the program recorded in the memory 704, and the voice feature storage unit. 102 is realized by storing the voice feature amount acquired by the voice feature extraction unit 101 in the memory 704. Further, the speaker / speech speed correspondence storage unit 105 is realized by storing the speech speed conversion rate in the memory 704 in association with the registered voice feature.

  The operation of the communication apparatus having the speech speed conversion apparatus according to Embodiment 2 of the present invention configured as described above will be described. FIG. 11 is a flowchart showing an example of the procedure of the speech speed conversion method according to Embodiment 2 of the present invention. Here, as in FIG. 5, it is assumed that the call device is connected to another point where a plurality of speakers exist and is talking.

  First, when it is necessary for the user of the call device to specify the voice of the other party's speaker that he / she wants to hear by converting his / her speech speed (Yes in step 21), the speech of such speaker Meanwhile, the user presses the registration button 605 to instruct the call device 601 to register voice features.

  Subsequently, when the user designates an appropriate speech rate conversion rate via an input unit such as the operation button 607 (step 22), the speech rate conversion unit 104 converts the speech rate at the designated conversion rate (step 22). Step 23). The user listens to the converted voice and determines whether the speaking speed is appropriate (step 24). If the speaking speed is not appropriate (No in step 24), the user returns to step 22 and the speaking speed is appropriate. Until the above is reached, the above-described processing is repeatedly executed while changing the conversion rate of the speech speed.

  After that, or when the speech speed is appropriate (Yes in step 24), the speech feature extraction unit 101 calculates the current speech feature amount, stores it in the speech feature storage unit 102, and is set The speech rate conversion rate is associated with the voice feature amount and stored in the speaker / speech rate correspondence storage unit 105 (step 25). The form of realizing the voice feature amount is the same as that described in the first embodiment.

  After that, or in the case where the speaker who is speaking at step 21 is not registered as the other party who wants to hear by converting the speech speed (in the case of No at step 21), the user performs another registration of the speaker whose speech speed is to be converted. In the case (No in Step 26), the process returns to Step 21 and the above-described processing is repeated until all the speakers whose speech speed is to be converted are registered.

  Registration of the speaker whose speech speed is to be converted is completed by the user (Yes in step 26), and it is determined by the communication device 601 whether or not the slow playback button 606 has been pressed by the user during normal conversation. (Step 27). When the slow playback button 606 is pressed (Yes in Step 27), the voice feature extraction unit 101 of the call device 601 continues to calculate the voice feature amount of the received voice (Step 28).

  Thereafter, the speaker determination unit 103 calculates the distances between the speech feature amounts of the received speech calculated in step 28 and the speech feature amounts of all the speakers stored in the speech feature storage unit 102, and from the results. It is determined whether or not the speaker is one of the speakers registered in step 25 (step 29). That is, if the distance from the voice feature amount is smaller than the threshold (Yes in step 29), the speaker determining unit 103 determines that the speaker matches the speaker registered in step 25, and the speech speed selecting unit 106 loads the speech speed conversion rate in the speaker / speech speed correspondence storage unit 105 registered in association with the matched speaker (step 30). Then, the speech speed conversion unit 104 converts the received voice at the speech speed conversion rate loaded by the speech speed selection unit 106 and reproduces it (step 31).

  On the other hand, when the distance from the voice feature amount is equal to or greater than the threshold (No in Step 29), or when the slow play button 606 is not pressed (No in Step 27), the speaker determination unit 103. Is determined not to be a speaker designated by the user, and the speech speed selection unit 106 instructs the speech speed conversion unit 104 to reproduce the received voice at a normal speed. Then, the speech speed conversion unit 104 performs a predetermined process without converting the speech speed on the received voice. And the process of step 21-step 31 is repeatedly performed until a telephone call is complete | finished. By performing such processing, it is possible to always apply good speech speed conversion with a conversion rate set by the user for each speaker regardless of which speaker on the other side of the call speaks.

  FIG. 12 is a conceptual diagram of real-time speech speed conversion by speaking speed settings of a plurality of speakers in Embodiment 2 of the present invention. In FIG. 12, with respect to the speaking speed of a certain speaker, the listener (speaker 1, speaker) is 110% for speaker 1, 120% for speaker 2, 130% for speaker 3, and so on. 2, speakers 3,...) Are registered with different speaking speeds. As shown in this figure, the voiced speech rate is expanded in the time direction with the conversion rate set for each listener, and the silent interval is compressed according to the actual speech timing, thereby converting the speech rate conversion rate. Even if they are different, speech speed conversion can be performed without causing a large delay that hinders conversation.

  Next, a specific method for realizing the speech speed conversion process according to the second embodiment will be described with reference to FIGS. FIG. 13 is a diagram showing an example of a configuration when using the communication device according to the second embodiment of the present invention, and FIG. 14 is a diagram showing an actual operation example of the speech speed conversion processing in FIG. In FIG. 13, 601 c to 601 f are call devices used by the speakers C to F, respectively, and 1201 c to 1201 f are gateways that connect the call devices 601 c to 601 f and the Internet 1202, respectively. In addition, the same code | symbol is attached | subjected to the component same as what was used in FIG. 7 and FIG. 8 of Embodiment 1, and the description is abbreviate | omitted.

  In the example of FIG. 13, it is assumed that a call is made between the four call devices 601c to 601f. In the case of the configuration of FIG. 13, the audio signal transmitted / received between the communication devices 601c to 601f is data in which a digital signal is packetized.

  Of course, the configuration at the time of use in the second embodiment is not limited to this, and as shown in the first embodiment, the communication devices 601c to 601f can be connected by an arbitrary method. Since this method has already been described in Embodiment 1, it is omitted here. In addition, the microphone 708 and the speaker 707 of the communication devices 601c to 601f according to the second embodiment are also structured so as not to generate howling, as in the first embodiment.

  An example of actual speech speed conversion processing in the above configuration will be described with reference to FIG. Here, the speaker C uses the communication device 601c of FIG. 13, the speaker D uses the communication device 601d, the speaker E uses the communication device 601e, and the speaker F uses the communication device 601f. Further, before starting the operation of FIG. 14, only the speaker F presses the registration button (corresponding to the registration button 605 of the call device 601 in FIG. 2) as described above, and steps 21 to step in the flowchart of FIG. 26, the speech rate conversion rates and voice characteristics of the speakers C to E are registered to the communication device 601f, and a slow playback button (corresponding to the slow playback button 606 of the communication device 601 in FIG. 2) is pressed. It is assumed that the speech device 601f has performed calculation of voice features, identification of a speaker, and selection of speech speed.

  Although omitted in FIG. 14, the speed of the voice spoken from the speaker F to the speakers C to E in the call device 601 f and the speaker F heard by the speakers C to E from the call devices 601 c to 601 e, respectively. The voice speeds are the same. As in 1411 and 1413 in FIG. 9 of the first embodiment, whether or not the speech rate conversion rate and the voice feature of the speaker F are registered in the communication devices 601c to 601e on the speaker C to E side. Or, even if registered, the slow playback button 606 of the talking devices 601c to 601e on the speakers C to E side has not been pressed, and the speech features are calculated and the speaker is identified and talked in the talking devices 601c to 601e. This is because the speed is not selected.

  In the former case, the processing state of each program shown in the flowchart of FIG. 11 installed in the communication devices 601c to 601e has not registered voice characteristics at the time of determination, so step 21 (Yes) → step 26 (Yes) → Control goes to step 27. Therefore, if the slow playback button 606 has been pressed, the process proceeds in the order of step 27 (Yes) → step 28 → step 29. In step 29, each speaker determination unit 103 of the communication devices 601 c to 601 e determines “No” based on the signal from each voice feature extraction unit 101, so that the determination result is received via each speech speed selection unit 106. The speech speed conversion units 104 of the communication devices 601c to 601e do not perform the speech speed conversion of the audio signal transmitted from the communication device 601f. Thereafter, the processing state of each program for executing the flowchart of FIG. 11 installed in the communication devices 601c to 601e shifts from step 29 (No) to step 21 respectively.

  In the latter case, since the voice feature is registered at the time of determination, the processing state of each program shown in the flowchart of FIG. 11 installed in the communication devices 601c to 601e is first step 21 (No) → step 26. (Yes) → Transition to Step 27. However, since the slow playback button 606 has not been pressed, “No” is determined in step 27 and the process proceeds to step 21. Each of the programs shown in the flowchart of FIG. 11 installed in the communication devices 601c to 601e is repeatedly performed. In the communication devices 601c to 601e, it is not necessary to operate the steps of each program shown in the flowchart of FIG.

  On the other hand, the voices 1621 to 1623 of the speakers C to E that are heard by the speaker F from the communication device 601f are talking to the speaker F from the speakers C to E in the communication devices 601c to 601e. The speed is slower than that of each of the voices 1611 to 1613. The voice speeds of the speakers C to E are set at the registration stage (steps 21 to 26) of the speaker (listener) F, respectively. This is because the voice speed and voice characteristics of the speakers C to E are registered in the call device 601f on the speaker F side, and the slow playback button 606 of the call device 601f on the speaker F side is pressed, This is because in the apparatus 601f, the calculation of the voice feature amount, the identification of the speaker, and the selection of the speech speed are performed.

  This processing state will be described with reference to the flowchart shown in FIG. When none of the speakers C to E is speaking to the speaker F, the processing state of the program shown in the flowchart of FIG. 11 installed in the communication device 601f is not registered as a voice feature at the time of determination. The process proceeds from step 21 (Yes) → step 26 (Yes) → step 27. Further, since the slow playback button 606 has been pressed, the process proceeds from step 27 (Yes) to step 29. In step 29, the speaker determination unit 103 in FIG. 4 determines “No” as a result of comparing the signal from the speech feature extraction unit 101, the signal from the speech feature extraction unit 101, and the data in the speech feature storage unit 102. The speech speed conversion unit 104 that has received the determination result does not perform the speech speed conversion of the voice signals transmitted from the communication devices 601c to 601e.

  Thereafter, the processing state of the program shown in the flowchart of FIG. 11 installed in the communication device 601f shifts from step 29 (No) to step 21. The state transition shown above is repeatedly performed by the program shown in the flowchart of FIG. 11 installed in the communication device 601f.

  On the other hand, for example, when speaker C is talking to speaker F (when in a state shown by section 1601 in FIG. 14), that is, voice 1611 in FIG. 14 is transmitted from communication device 601c to communication device 601f. 11, the processing state of the program shown in the flowchart of FIG. 11 installed in the communication device 601 f does not register voice characteristics at the time of determination, so step 21 (Yes) → step 26 (Yes) → step 27. Migrate to

  Further, since the slow playback button 606 has been pressed, the processing state of the program shown in the flowchart of FIG. 11 installed in the call device 601f is determined as “Yes” in step 27, and step 28 → step 29 Move in order. As a result of comparing the signal from the voice feature extraction unit 101 and the data in the voice feature storage unit 102 by the speaker determination unit 103 of the telephone conversation device 601f in step 29, “speaker C”, that is, “Yes” in step 29 is obtained. The state transitions in the order of step 30 → step 31. That is, the speech speed conversion unit 104 of the communication device 601f that has received the determination result of “Yes” in step 29, in step 30, the speaker C from the speaker / speech speed correspondence storage unit 105 via the speech speed selection unit 106. The speech rate conversion rate (110%) corresponding to is loaded, and the speech rate of the voice signal 1611 of the speaker C sent from the communication device 601c is converted in step 31 to 1621. Thereafter, the processing state of the program shown in the flowchart of FIG. 11 installed in the communication device 601f shifts from step 31 to step 21. The processing described above is performed in the same manner for the voices 1612 and 1613 of the speakers D and E, and for the speaker F, the playback voice 1622 converted to 120% and the playback voice 1623 converted to 130%, respectively. Sounds like this.

  In the second embodiment, the case where only the speaker F presses the slow playback button 606 after registering the speech rate conversion rate and voice characteristics of the other speakers C to E has been described. However, not only the speaker F but also other speakers C to E may perform the above-described speech rate conversion rate, voice feature registration, and slow playback button 606 pressing, and a plurality of speakers may be simultaneously active. You may implement. Further, two or three communication devices 601 may be used, or five or more communication devices 601 may be used. When a plurality of speakers speak in one communication device 601, the other communication devices 601 are connected to the plurality of speakers. It is also possible to register speech characteristics and speech rate conversion rate and perform speech rate conversion suitable for each speaker.

  As described above, in the second embodiment, the speech rate conversion rate set corresponding to the speaker storing the voice feature is stored in the speaker / speech rate correspondence storage unit 105, and the speech rate selection unit 106 Since the speech rate conversion rate corresponding to the speaker determined by the determination unit 103 is extracted, and the speech rate conversion unit 104 converts the received voice based on this, in a call involving a plurality of speakers, It is possible to listen to the content of another speaker's speech at a speed according to the listener's sense, and the intelligibility is improved. On the other hand, since speech speed conversion is not applied to the voice of a speaker not designated by the listener, it is possible to avoid the conversion from having an adverse effect and lowering the intelligibility of the conversation.

(Embodiment 3)
FIG. 15 is a block diagram schematically showing a configuration of a portion related to the speech rate conversion apparatus according to Embodiment 3 of the present invention. In FIG. 15, as in the first embodiment, 101 is a voice feature extraction unit that extracts voice features of individual speakers participating in a call, 102 is a voice feature storage unit that stores the extracted voice features, and 103. Determines whether the speaker is one of the speakers designated by the listener by comparing the voice of the current speaker with the voice features stored in the voice feature storage unit 102. Reference numeral 104 denotes a speech speed conversion unit that receives the determination result of the speaker determination unit 103 and converts the speech speed of the received voice.

  Reference numeral 107 denotes a speaker / speech speed / volume correspondence storage section for storing an optimum speaking speed conversion rate and reproduction volume amplification ratio set for each speaker, and 108 denotes a speaker determination section 103. This is a speech speed / volume selector that selects the conversion rate of the speech speed and the amplification factor of the playback volume corresponding to the determined speaker from the speaker / speech speed / volume correspondence storage unit 107. Here, the speech feature stored in the speech feature storage unit 102, the conversion rate of the speech speed and the amplification factor of the playback volume stored in the speaker / speech speed / volume correspondence storage unit 107 are, for example, the target of the speech feature Are associated by speaker identification information attached to the speaker. The speech speed conversion unit 104 converts the speech speed and playback volume of the corresponding speaker based on the speech speed conversion rate selected by the speech speed / volume selection unit 108 and the playback volume amplification rate. The designated speaker conversion condition storage means in the claims corresponds to the speaker / speech speed / volume correspondence storage section 107, and the designated speaker conversion condition selection means similarly corresponds to the speech speed / volume selection section 108. .

  With such a configuration, the speech speed conversion unit 104 converts the speech speed at the conversion rate received from the speech speed / volume selection unit 108 and at the same time converts the volume so that the speech distance and the loudness of the speaker can be adjusted. Regardless, the voice of each speaker is automatically converted to the listener's preferred volume and speed. Note that the configuration of the communication device including the speech speed conversion device according to the third embodiment is the same as that shown in FIGS. 1 to 3 of the first embodiment, and thus the description thereof is omitted.

  In the third embodiment, the voice feature storage unit 102 and the speaker / speech speed / volume correspondence storage unit 107 correspond to the memory 704 in FIG. Further, the speaker determination unit 103, the speech speed conversion unit 104, and the speech speed / volume selection unit 108 are stored in the memory 704 as software programs that operate on the DSP 701 in FIG. That is, the speaker determination unit 103, the speech speed conversion unit 104, and the speech speed / volume selection unit 108 calculate the received voice received by the DSP 701 from the communication path interface 702 according to the program stored in the memory 704 and perform D / This is realized by outputting to the A converter 705. The voice feature extraction unit 101 is realized by calculating the voice feature amount from the received voice received by the DSP 701 from the communication path interface 702 according to the program recorded in the memory 704, and the voice feature storage unit. 102 is realized by storing the voice feature amount acquired by the voice feature extraction unit 101 in the memory 704. Further, the speaker / speech speed / volume correspondence storage unit 107 stores the speech rate conversion rate converted by the speech rate conversion unit 104 and the reproduction volume amplification rate in the memory 704 in association with the registered voice feature. It is realized with.

  The operation of the communication apparatus having the speech rate conversion apparatus according to Embodiment 3 of the present invention configured as described above will be described. FIG. 16 is a flowchart showing an example of the procedure of the speech speed conversion method according to Embodiment 3 of the present invention. Here, similarly to FIG. 5 of the first embodiment and FIG. 11 of the second embodiment, it is assumed that the call device is connected to another point where a plurality of speakers exist and is in a call.

  Similar to steps 21 to 24 in FIG. 11 of the second embodiment, the user of the communication device 601 designates the speech rate conversion rate of the voice of the other party's speaker that he / she wants to hear by converting the speech rate ( Step 41 to Step 44). Thereafter, when the user designates an appropriate volume (step 45), the speech speed conversion unit 104 amplifies the designated volume. The user listens to the amplified voice and determines whether the sound volume is appropriate (step 46). If the sound volume is not appropriate (No in step 46), the user returns to step 45 again to determine that the sound volume is appropriate. Until it is, the above-described processing is repeatedly executed while changing the reception volume.

  If the sound volume is appropriate (Yes in step 46), the speech feature extraction unit 101 calculates the current speech feature amount, stores it in the speech feature storage unit 102, and sets the set speech rate conversion rate. Are stored in the speaker / speech speed / volume correspondence storage unit 107 in association with the voice feature amount (step 47). The form of realizing the voice feature amount is the same as that described in the first embodiment.

  After that, or when the speaker who is speaking in step 41 is not registered as the other party who wants to hear by converting the speech speed and / or volume (if No in step 41), the user can change the speech speed and / or volume. If the speaker to be converted is registered elsewhere (No in step 48), the process returns to step 41 until the above-described processing registers all the speakers for which the speech speed conversion and / or volume conversion is to be performed. Repeated.

  Determination of whether or not the slow playback button 606 has been pressed by the user during normal conversation after registration of the speaker who wants to convert the speech speed and / or volume is completed by the user (Yes in step 48). (Step 49). When the slow playback button 606 is pressed (Yes in Step 49), the voice feature extraction unit 101 of the call device 601 continues to calculate the voice feature amount of the received voice (Step 50).

  Thereafter, the speaker determination unit 103 calculates the distances from the voice feature amounts of all the speakers stored in the voice feature storage unit 102, and from the results, one of the speakers registered in step 47 is calculated. It is determined whether or not there is (step 51). That is, if the distance from the voice feature amount is smaller than the threshold value (Yes in step 51), the speech speed / volume selection unit 108 stores the speaker / speech speed / volume correspondence memory associated with the matched speaker. The speech rate conversion rate and the reproduction volume gain in the unit 107 are loaded (step 52). Then, the speech speed conversion unit 104 converts the received voice by the speech speed conversion rate loaded by the speech speed / volume selection unit 108 and the amplification factor of the playback volume and reproduces it (step 53).

  On the other hand, if the distance from the voice feature amount is equal to or greater than the threshold (No in Step 51) or if the slow playback button 606 is not pressed (No in Step 49), the speaker determination unit 103 The speech speed / volume selection unit 108 instructs the speech speed conversion unit 104 to reproduce the received voice at a normal speed and volume. Then, the speech speed conversion unit 104 performs a predetermined process without converting the speech speed or the volume of the received voice. And the process of step 41-step 53 is repeatedly performed until a telephone call is complete | finished. By performing such processing, regardless of the other party's speaker who speaks, the conversion rate and playback volume conversion rate set by the user for each speaker can always be used to achieve good speech speed conversion and playback processing. It can be carried out.

  Next, a specific method of realizing the speech speed conversion process in the third embodiment will be described with reference to FIGS. FIG. 17 is a diagram showing an actual operation example of the speech speed conversion process in FIG. Here, the case where it has the same structure as FIG. 13 in Embodiment 2 is assumed. That is, it is a case where a call is made between the four call devices 601c to 601f. The communication devices 601c to 601f are connected to the Internet 1202 via gateways 1201c to 1201f, respectively. Further, it is assumed that the audio signal transmitted and received between the communication devices 601c to 601f is data obtained by packetizing a digital signal. The configuration in use in the third embodiment is not limited to this, and there are variations similar to those described in the first and second embodiments.

  An example of actual speech speed conversion processing in the above configuration will be described with reference to FIG. Here, as in FIG. 14 of the second embodiment, the speaker C uses the communication device 601c of FIG. 13, the speaker D uses the communication device 601d, the speaker E uses the communication device 601e, The speaker F uses the communication device 601f. Also, before starting the operation of FIG. 17, only the speaker F presses the registration button (corresponding to the registration button 605 of the call device 601 in FIG. 2) as described above, and steps 41 to step in the flowchart of FIG. 48, the speech rate conversion rate, the reproduction volume amplification factor, and the voice characteristics of the speakers C to E are registered to the call device 601f, and a slow play button (to the slow play button 606 of the call device 601 in FIG. 2) is registered. It is assumed that the call feature 601f is calculated, the speaker is specified, the speaker is specified, and the speech speed and reception volume are selected.

  Although omitted in FIG. 17, the speed and volume of the voice talking from the speaker F to the speakers C to E in the call device 601 f and the speakers C to E are heard from the call devices 601 c to 601 e, respectively. The speed and volume of the voice of the speaker F are the same. Similarly to 1411 and 1413 in FIG. 9 of the first embodiment, the speech rate conversion rate, volume, and voice characteristics of speaker F are not registered in the communication devices 601c to 601e on the speakers C to E side. Or, even if registered, the slow playback button 606 of the talking devices 601c to 601e on the speakers C to E side is not pressed, and the speech features are calculated and the speaker is specified in the talking devices 601c to 601e. This is because the volume and speaking speed are not selected.

  In the former case, the processing state of each program shown in the flowchart of FIG. 16 installed in the communication devices 601c to 601e has not registered voice characteristics at the time of determination, so step 41 (Yes) → step 48 (Yes) → Control goes to step 49. Therefore, if the slow playback button 606 has been pressed, the process proceeds in the order of step 49 (Yes) → step 50 → step 51. In step 51, each speaker determination unit 103 of the call devices 601 c to 601 e determines “No” based on the signal from each voice feature extraction unit 101, and therefore the determination result via each speech speed / volume selection unit 108. The speech rate conversion units 104 of the call devices 601c to 601e that have received the call do not perform the speech rate conversion of the voice signal transmitted from the call device 601f. Thereafter, the processing state of each program shown in the flowchart of FIG. 16 installed in the communication devices 601c to 601e shifts from step 51 (No) to step 41, respectively.

  In the latter case, since the voice feature is registered at the time of determination, the processing state of each program shown in the flowchart of FIG. 16 installed in the communication devices 601c to 601e is first step 41 (No) → step 48. (Yes) → Transition to step 49. However, since the slow playback button 606 has not been pressed, “No” is determined in step 49 and the process proceeds to step 41. Each of the programs shown in the flowchart of FIG. 11 installed in the communication devices 601c to 601e is repeatedly performed. In the communication devices 601c to 601e, it is not necessary to operate the steps of the programs shown in the flowchart of FIG.

  On the other hand, the voices 1721 to 1723 of the speakers C to E that are heard by the speaker F from the call device 601f are the voices that are being spoken to the speaker F from the speakers C to E in the call devices 601c to 601e. The speed is slower than 1711 to 1713, and the respective volumes are also optimum. The voice speed and volume of each speaker C to E are set in the registration stage (steps 41 to 48) of the speaker (listener) F, respectively. This is because the voice speed, volume and voice characteristics of the speakers C to E are registered in the call device 601f on the speaker F side, and the slow playback button 606 of the call device 601f on the speaker F side is pressed. This is because the voice feature amount is calculated, the speaker is specified, and the speech speed and volume are selected in the call device 601f. Note that the size of the substantially rectangular audio signal in FIG. 17 in the horizontal direction (left-right direction in the figure) indicates the speed of speech, and the longer the length, the slower the speech speed. Further, the size of the substantially rectangular audio signal in the vertical direction (vertical direction in the figure) indicates the volume, and the longer the volume, the higher the volume.

  This processing state will be described with reference to the flowchart shown in FIG. When none of the speakers C to E is speaking to the speaker F, the processing state of the program shown in the flowchart of FIG. The process proceeds from step 41 (Yes) to step 48 (Yes) to step 49. Since the slow playback button 606 has been pressed, the processing state of the program shown in the flowchart of FIG. Move in order. In step 51, the speaker determination unit 103 of the call device 601f determines “No” as a result of comparing the signal from the voice feature extraction unit 101 and the data of the voice feature storage unit 102. Unit 104 does not perform speech speed conversion or volume conversion on the audio signals sent from communication devices 601c to 601e.

  Thereafter, the processing state of the program shown in the flowchart of FIG. 16 installed in the communication device 601f shifts from step 51 (No) to step 41. The state transition shown above is repeatedly performed by the program shown in the flowchart of FIG. 16 installed in the communication device 601f.

  On the other hand, for example, when speaker C is speaking to speaker F (when in a state shown by section 1601 in FIG. 17), that is, voice 1711 in FIG. 17 is transmitted from communication device 601c to communication device 601f. 16, the processing state of the program shown in the flowchart of FIG. 16 installed in the communication device 601 f is not registered at the time of determination, so that the voice feature is not registered, so step 41 (Yes) → step 48 (Yes) → step 49 ( Yes).

  Since the slow playback button 606 has been pressed, the processing state of the program shown in the flowchart of FIG. 16 installed in the call device 601f is determined to be “Yes” in step 49, and from step 50 to step 51. Move in order. As a result of comparison of the signal from the voice feature extraction unit 101 and the data of the voice feature storage unit 102 by the speaker determination unit 103 of the call device 601f in step 51, “speaker C”, that is, “Yes” in step 51 is obtained. The state transitions in the order of step 52 to step 53. In other words, the speech speed conversion unit 104 of the call device 601f that has received the determination result of “Yes” in step 51 passes the talk speed / volume selection unit 108 from the speaker / speech speed / volume correspondence storage unit 107 in step 52. Then, the speech rate conversion rate and the reproduction volume amplification factor of the speaker C are loaded, and the speech speed and playback volume conversion of the speech signal 1711 of the speaker C sent from the call device 601c are performed in step 53, and 1721 To do. The processing described above is performed in the same manner for the voices 1712 and 1713 of the speakers D and E, so that the speaker F can be heard as reproduced voices 1722 and 1723, respectively.

  In the third embodiment, the case where only the speaker F presses the slow playback button 606 after registering the speech rate conversion rate and voice characteristics of the other speakers C to E has been described. However, the speech rate conversion rate, the reproduction volume amplification factor, the voice feature registration, and the slow playback button 606 are not limited to the speaker F, but may be performed by other speakers C to E. A plurality of speakers may be performed simultaneously. Further, two or three communication devices 601 may be used, or five or more communication devices 601 may be used. When a plurality of speakers speak in one communication device 601, the other communication devices 601 are connected to the plurality of speakers. It is also possible to register the voice feature, the speech speed conversion rate, and the reproduction volume amplification rate, and perform the speech speed conversion and playback volume conversion suitable for each speaker.

  As described above, in the third embodiment, the speech rate conversion rate and the speech amplification rate set for the speaker storing the speech features are stored in the speaker / speech rate / volume correspondence storage unit 107, and the speech The speed / volume selection unit 108 extracts the speech rate conversion rate and the speech amplification rate corresponding to the speaker determined by the speaker determination unit 103, and the speech rate conversion unit 104 converts the received speech based on this. Therefore, in a call in which a plurality of speakers participate, it is possible to hear the utterance contents of other speakers at a speed and volume according to the listener's sense, and the intelligibility is improved. On the other hand, since the speech speed conversion is not applied to the voice of the speaker not specified by the listener, it is also possible to avoid further increase / decrease in volume and decrease in the intelligibility of the conversation due to the reverse effect of the conversion. it can.

  As described above, the speech rate conversion device, the speech device, and the speech rate conversion method according to the present invention are useful for a conference call system such as a conference call in which a plurality of speakers are talked to each other.

The perspective view which shows an example of the communication apparatus provided with the speech speed conversion apparatus in Embodiment 1 of this invention. Top view of communication apparatus according to Embodiment 1 of the present invention. The block diagram which shows typically the structure of the component in which the speech-speed converter by Embodiment 1 of this invention is related with the telephone call apparatus in Embodiment 1 of this invention. The block diagram which shows typically the function structure of the speech-speed converter in Embodiment 1 of this invention. The flowchart which shows an example of the procedure of the speech speed conversion method in Embodiment 1 of this invention. The figure which shows the operation | movement concept of the speech speed conversion process in Embodiment 1 of this invention. The figure which shows an example of a structure at the time of use of the telephone call apparatus in Embodiment 1 of this invention. The figure which shows the other example of a structure at the time of use of the telephone call apparatus in Embodiment 1 of this invention. The figure which shows the actual operation example of the speech speed conversion process in Embodiment 1 of this invention. The block diagram which shows typically the structure of the part with which the speech-speed converter in Embodiment 2 of this invention is related. The flowchart which shows an example of the procedure of the speech speed conversion method in Embodiment 2 of this invention. Conceptual diagram of real-time speech speed conversion by speaking speed setting of a plurality of speakers in Embodiment 2 of the present invention The figure which shows an example of the structure at the time of use of the telephone call apparatus in Embodiment 2 of this invention. The figure which shows the actual operation example of the speech speed conversion process in Embodiment 2 of this invention. The block diagram which shows typically the structure of the part to which the speech-speed converter in Embodiment 3 of this invention is related. The flowchart which shows an example of the procedure of the speech speed conversion method in Embodiment 3 of this invention. The figure which shows the actual operation example of the speech speed conversion process in Embodiment 3 of this invention. A block diagram schematically showing the configuration of a conventional speech speed converter

Explanation of symbols

DESCRIPTION OF SYMBOLS 101 Voice feature extraction part 102 Voice feature memory | storage part 103 Speaker determination part 104 Speak speed conversion part 105 Speaker / speech speed correspondence memory | storage part 106 Speak speed selection part 107 Speaker / speech speed / volume correspondence memory | storage part 108 Selection unit 601, 601 a to 601 f, communication device 602 a to 602 d, 708 microphone 603, 707 speaker 605 registration button 606 slow playback button 607 operation button 702 communication path interface 704 memory 705 D / A converter 706 A / D converter 1201 a to 1201 f gateway 1202 Internet 1301a, 1301b Connection line 1302a, 1302b Modem 1303a, 1303b Public line network 1304a, 1304b Internet service provider

Claims (7)

Voice feature extraction means for extracting voice features of individual speakers participating in the call;
Voice feature storage means for storing voice features extracted for a speaker designated by the listener;
Talk to determine whether the speaker is one of the speakers specified by the listener by comparing the voice features of the current speaker with the voice features stored in the voice feature storage means Person determination means;
A speech speed converting means for converting the speech speed of the received voice of the speaker when the speaker determining means determines that the speaker is designated by the listener;
A speech speed conversion device comprising: Designated speaker conversion condition storage means for storing a conversion rate of the optimum speech speed set for each of the speakers specified by the listener;
A designated speaker conversion condition selecting means for selecting a conversion rate of the speech speed corresponding to the speaker determined by the speaker determining means from the specified speaker conversion condition storage means;
The speech speed conversion means converts the speech speed of the speaker specified by the listener using the conversion rate of the speech speed selected by the designated speaker conversion condition selection means. The speech speed conversion apparatus according to claim 1. The designated speaker conversion condition storage means further stores a reproduction volume amplification factor set for each of the speakers designated by the listener,
The designated speaker conversion condition selection means selects, from the designated speaker conversion condition storage means, the conversion rate of the speech speed and the amplification factor of the reproduction volume corresponding to the speaker determined by the speaker determination means,
The speaking speed conversion means uses the speaking speed conversion rate selected by the designated speaker conversion condition selection means and the reproduction volume amplification factor, and the speaking speed and reproduction of the speaker specified by the listener. The speech rate conversion apparatus according to claim 2, wherein the speech rate conversion unit converts the volume. A communication means for communicating with another communication device via a communication line;
A sound collection means for collecting the voice of the speaker;
Audio output means for reproducing and outputting audio from the other call device;
The speech rate conversion device according to any one of claims 1 to 3,
A call device comprising: Perform voice extraction processing to extract the voice features of individual speakers participating in the call, perform voice feature storage processing to store the voice features extracted for the speaker specified by the listener, and A speaker determination process is performed to determine whether or not the speaker is one of the speakers designated by the listener by comparing a feature with the stored voice feature. A speech speed conversion method, comprising: performing speech speed conversion processing for converting the speech speed of a received voice of the speaker when it is determined that the speaker is designated by a listener. Performing designated speaker conversion condition storage processing for storing the conversion rate of the optimum speech speed set for each of the speakers designated by the listener,
Specified speaker conversion condition selection for selecting the conversion rate of the speech speed corresponding to the speaker determined by the speaker determination process before the speech speed determination process from the contents stored in the specified speaker conversion condition storage process Process
The speech speed conversion process converts the speech speed of the speaker specified by the listener using the conversion rate of the speech speed selected in the designated speaker conversion condition selection process. 5. The speech speed conversion method according to 5. In the designated speaker conversion condition storage process, the reproduction volume amplification factor set for each of the speakers designated by the listener is further stored,
In the designated speaker conversion condition selection process, the content stored in the designated speaker conversion condition storage process, the conversion rate of the speech speed and the amplification factor of the reproduction volume corresponding to the speaker determined by the speaker determination process Select from
In the speech speed conversion process, the speech speed and playback of the speaker specified by the listener using the conversion rate of the speech speed selected in the designated speaker conversion condition selection process and the amplification factor of the playback volume. The speech speed converting method according to claim 6, wherein the sound volume is converted.

Images