JP2017216602A - Telephone device - Google Patents

Abstract

PROBLEM TO BE SOLVED: To efficiently perform speech speed conversion that is easy to hear for a user, while suppressing erroneous detection of a voice interval included in a speed reception signal.SOLUTION: A microphone 11 collects speech transmission voices of a user. A noise level estimation part 21z estimates a level of a noise signal around the user. A voice internal detection part 21 detects an interval of a voice included in a transmission signal from the other telephone set. While the speech transmission voices of the user are collected by the microphone 11, a speech reception gain control part 15 attenuates a speech reception signal including a line echo component of the speech transmission voice based on a telephone switching network. Based on an attenuation amount of the speech reception signal and the estimated level of the noise signal, an interval detection correction part 17 corrects a threshold value to be used for the voice interval detection in the voice interval detection part 21. Based on the threshold that is corrected by the interval detection correction part 17, a voice conversion part 22 converts a speech speed of the voice included in the transmission signal from the other telephone set and outputs the voice from a speaker 12.SELECTED DRAWING: Figure 2

Description

  The present invention relates to a telephone device that receives a voice signal of a call transmitted from a telephone device of a call partner, converts a speech speed of a voice section of the voice signal, and outputs the converted voice speed.

  The speech speed conversion device, for example, expands a voice signal of a call transmitted from the telephone device of the other party (that is, the voice signal on the receiving side) at a constant rate in the time direction, so that the other party speaks slowly and listens to the listener. Has a function to convert the sound into a voice that makes it easier to hear.

  This speech speed conversion device is not limited to a telephone device. As a prior art using a speech speed conversion device, for example, a speech speed conversion device mounted on an interphone is known (see Patent Document 1). The speech speed conversion apparatus of Patent Document 1 discriminates between a speech section in which speech is included in an input signal and a non-speech section in which speech is not included, and performs expansion processing when it is determined as a speech section, Compression processing is performed when it is determined as a non-voice section. In addition, even when the speech speed converting apparatus determines that it is a non-speech segment, it does not perform compression processing when the noise level included in the input signal is equal to or higher than a predetermined threshold value. Thereby, the speech speed converting apparatus can prevent the output voice from being interrupted due to the input voice being compressed by mistake.

Japanese Patent No. 5346230

  However, when the configuration of the speech speed conversion device described in Patent Document 1 is applied to a telephone device used for real-time phone calls, the following problems have been considered. In a real-time call, conversation is generally performed in a time-sharing manner between the user's own telephone device and the telephone device of the other party. That is, when the user is talking, the other party is listening to the conversation, and when the other party is talking, the user is listening to the conversation. In transmission / reception of call voice between telephone devices during such a real-time call, a line echo based on a two-wire / four-wire conversion circuit generally provided in a telephone exchange network occurs. For this reason, for example, the line echo component of the voice of the user himself (that is, the transmitted voice) may enter the receiver circuit of the telephone device, and the line echo component may be heard from a speaker provided in the receiver circuit. As a result, it is conceivable that the line echo component becomes noise and hinders a call.

  Further, in order to suppress the generation of noise due to the line echo component, for example, if a voice switch is mounted on the telephone device, it is possible to reduce the received signal when the user of the telephone device transmits. However, the level of the noise signal included in the received voice varies depending on whether the voice switch is turned on or off. As a result, the speech speed conversion device provided in the telephone device may erroneously detect a fluctuation in the level of the noise signal as voice. For example, since the voice switch stops the attenuation of the received signal when the user's transmission is completed, the speech speed converting device uses the voice switch to turn off the voice switch, and the level of noise due to the line echo component increases. Since it is determined that the signal is a signal, the speech section cannot be normally detected (see FIG. 7B). FIG. 7B is a timing chart showing changes in the received voice before and after the voice switch is turned on.

  Therefore, if the speech rate conversion device erroneously detects the speech interval, the speech rate conversion cannot be performed properly for the speech interval included in the received signal, and noise in the non-speech interval is also converted to the speech rate. Become. As a result, there is a time unnecessary for the speech speed conversion, and it becomes difficult to hear the received speech, and the effect of the speech speed conversion is reduced. There was also an effect that the sound was interrupted.

  The present invention has been made in view of the above-described conventional situation, and provides a telephone device that can suppress erroneous detection of a voice section included in a received signal and can efficiently perform speech speed conversion that is easy for a user to hear. The purpose is to provide.

  The present invention relates to a telephone device for making a call with another telephone device via a telephone exchange network, a microphone for collecting a user's transmitted voice, and a noise for estimating a level of a noise signal around the user. Based on the telephone switching network while the level estimation unit, the section detection unit for detecting the voice section included in the transmission signal from the other telephone device, and the user's transmitted voice is collected by the microphone A reception gain control unit for attenuating a reception signal including a line echo component of the transmission voice; and an attenuation amount of the reception signal by the reception gain control unit and a level of the noise signal estimated by the noise level estimation unit. Based on the threshold correction unit that corrects the threshold value used for the voice segment detection in the segment detection unit and the threshold value corrected by the threshold correction unit, it is included in the transmission signal from the other telephone device. Comprising a speech rate conversion section outputting from the speaker to the speech speed converting speech, and a telephone device.

  ADVANTAGE OF THE INVENTION According to this invention, the false detection of the audio | voice area contained in a received signal can be suppressed, and speech speed conversion which is easy to hear for a user can be performed efficiently.

The figure which shows the connection between the telephone sets in 1st Embodiment The figure which shows in detail an example of the internal structure of the part in connection with the speech speed conversion of the telephone connected to a 2-wire 4-wire conversion circuit Block diagram showing in detail an example of the internal configuration of the speech rate conversion unit (A), (B) Timing chart showing audio signals before and after speech speed conversion (A) Timing chart showing detection result of voice section, (B) Timing chart showing time average value of signal waveform, (C) Timing chart of voice waveform The flowchart explaining in detail an example of a call operation procedure of the first embodiment (A), (B) Timing chart showing changes in received voice before and after the voice switch is turned on (A), (B) The figure explaining the method of the speech speed conversion with respect to the sound containing a vowel and a consonant (A), (B) Timing chart showing time change of AMDF value The figure which shows an example of the internal structure of the part in connection with speech speed conversion in the telephone of 2nd Embodiment in detail Timing chart showing an example of state change in each part of the telephone A flowchart for explaining an example of a call / playback operation procedure according to the second embodiment

  Hereinafter, embodiments of a telephone device according to the present invention will be described with reference to the drawings.

(Background and issues leading to the first embodiment)
The speech speed conversion device has a function of converting the original voice into a voice similar to that when the speaker slowly speaks by expanding the voice signal on the receiver side in the time direction. When the configuration of the speech speed conversion device described in Patent Document 1 is applied to a telephone device used for real-time phone calls, the following problems can be considered. In a real-time call, conversation is generally performed in a time-sharing manner between the user's own telephone device and the telephone device of the other party. That is, when the user is talking, the other party is listening to the conversation, and when the other party is talking, the user is listening to the conversation. In the transmission / reception of call voice between telephone devices during such a real-time call, a line echo based on a two-wire / four-wire conversion circuit provided in a telephone is generally generated. For this reason, for example, the line echo component of the voice of the user himself (that is, the transmitted voice) may enter the receiver circuit of the telephone device, and the line echo component may be heard from a speaker provided in the receiver circuit. As a result, it is conceivable that the line echo component becomes noise and hinders a call.

  In addition, in order to suppress the generation of noise due to the line echo component, for example, when a voice switch is mounted on the telephone device, the received signal including voice and noise signals is reduced when the user of the telephone device transmits. It is possible. However, the level of the noise signal included in the received voice varies depending on whether the voice switch is turned on or off. As a result, the speech speed conversion device provided in the telephone device may erroneously detect a fluctuation in the level of the noise signal as voice. For example, since the voice switch stops the attenuation of the received signal when the user's transmission is completed, the speech speed converting device uses the voice switch to turn off the voice switch, and the level of noise due to the line echo component increases. Since it is determined that the signal is a signal, the speech section cannot be normally detected (see FIG. 7B).

  Therefore, if the speech rate conversion device erroneously detects the speech interval, the speech rate conversion cannot be performed properly for the speech interval included in the received signal, and noise in the non-speech interval is also converted to the speech rate. Become. As a result, there is a time unnecessary for the speech speed conversion, and it becomes difficult to hear the received speech, and the effect of the speech speed conversion is reduced. There was also an effect that the sound was interrupted.

  Therefore, in the first embodiment, an example of a telephone device that can suppress erroneous detection of a voice section included in a received signal and can efficiently perform speech speed conversion that is easy for a user to hear will be described.

  Hereinafter, embodiments that specifically disclose a transfer telephone device according to the present invention will be described in detail with reference to the drawings as appropriate. However, more detailed description than necessary may be omitted. For example, detailed descriptions of already well-known matters and repeated descriptions for substantially the same configuration may be omitted. This is to avoid the following description from becoming unnecessarily redundant and to facilitate understanding by those skilled in the art. The accompanying drawings and the following description are provided to enable those skilled in the art to fully understand the present disclosure, and are not intended to limit the subject matter described in the claims.

(First embodiment)
FIG. 1 is a diagram illustrating a connection between the telephones 10A and 10B according to the first embodiment. A telephone 10A and a telephone 10B as an example of a telephone device according to the present invention are connected to each other via a telephone exchange network 7 including a public analog line 50 so as to be able to make a call.

  The telephone 10A has a two-wire four-wire conversion circuit 30A that serves as an exchange. The two-wire four-wire conversion circuit 30A converts a total of four signals for every two wires into a two-wire signal of a ground and a signal wire with respect to the microphone and the speaker (described later) in the telephone 10A. To do. The 2-wire 4-wire conversion circuit 30A is connected to the public analog line 50.

  Similarly, the telephone set 10B includes a two-wire / four-wire conversion circuit 30B that serves as an exchange. The two-wire / four-wire conversion circuit 30B converts a total of four signals for every two wires, plus and minus, with respect to the microphone and the speaker (see later) in the telephone 10B, into two signals of the ground and the signal wire. To do. The 2-wire 4-wire conversion circuit 30B is connected to the public analog line 50. The two-wire / four-wire conversion circuits 30A and 30B are connected to the public analog line 50 by two wires.

  The telephones 10A and 10B of the present embodiment are fixed telephones that are connected to a public analog line 50 that is generally popular.

  Note that the telephones 10A and 10B may be simply referred to as the telephone 10 when it is not necessary to distinguish between them. In addition, when there is no need to distinguish between the two-wire four-wire conversion circuits 30A and 30B, they may be simply referred to as the two-wire four-wire conversion circuit 30.

  In addition, the exchange described above may be a private branch exchange (PBX) that is installed in a company office or the like and exchanges an extension telephone and an external telephone, for example. In this case, a business phone connected to the private branch exchange is used as the telephone. In the present embodiment, it is assumed that a telephone call is made using a public analog line, but the present invention can be similarly applied to a telephone call using a digital line.

  FIG. 2 is a diagram showing in detail an example of the internal configuration of the portion related to the speech speed conversion of the telephone set 10 connected to the 2-wire 4-wire conversion circuit 30. The telephone 10 shown in FIG. 2 includes a microphone 11, a speaker 12, a reception gain control unit 15, a section detection correction unit 17, a slow talk button 16, and a speech speed conversion unit 20.

  The microphone 11 collects and inputs the voice of the user who uses the telephone 10 (that is, the voice on the transmission side).

  The speaker 12 outputs the voice (that is, the voice on the receiving side) included in the transmission signal from the telephone 10 of the other party.

  The reception gain control unit 15 uses, for example, any one of a voice switch, a center clipper, and an AGC (Auto Gain Control) to attenuate the level of the reception signal of the telephone 10 when the user's voice is transmitted. To control. The received signal includes a signal of a line echo component based on the two-wire / four-wire conversion circuit 30, and further a voice or noise signal.

  For example, the voice switch determines whether or not there is a transmission in the telephone 10 and performs switching of transmission and reception. When it is determined that there is a transmission, the level of the reception signal is attenuated.

  The center clipper sets the level of the received signal to a value of about 0 when the telephone 10 transmits.

  When the voice on the transmitting side of the telephone 10 is high, the AGC lowers the level of the received signal according to the volume.

  The reception gain control unit 15 outputs the reception side attenuation to the section detection correction unit 17 as needed.

  The section detection correction unit 17 as an example of the threshold correction unit corrects a threshold used for speech section detection by the speech speed conversion unit 20 based on information on the attenuation amount of the reception signal obtained from the reception gain control unit 15. . There are, for example, three methods for correcting the threshold value used for detecting the voice section. The threshold is set using a noise level that is, for example, a long-term average AvL (see FIG. 5B) of the received signal. The long-term average AvL of the received signal is obtained by being derived by the noise level estimation unit 21z.

  In the first correction method, the section detection correction unit 17 multiplies the long-term average AvL of the noise level estimated by the noise level estimation unit 21z by the attenuation amount (gain) of the reception obtained from the reception gain control unit 15. The threshold is set by combining them. That is, according to the first correction method, the threshold value is lowered according to the attenuation amount of the received signal.

  In the second correction method, the section detection correction unit 17 transmits a control signal indicating that the reception signal is amplified according to the attenuation while the reception signal is attenuated by the reception gain control unit 15. 21 is instructed. The signal amplification unit 21x of the speech section detection unit 21 amplifies the received reception signal in accordance with the control signal from the section detection correction unit 17, and estimates the noise level from the amplified reception signal. That is, according to the second correction method, the threshold is set according to the estimated noise level included in the amplified received signal.

  In the third correction method, the section detection correction unit 17 does not estimate the noise level by the noise level estimation unit 21z of the voice section detection unit 21 while the reception signal is attenuated by the reception gain control unit 15. That is, according to the third correction method, the threshold value is set to a predetermined noise level (that is, the long-term average AvL of the received signal shown in FIG. 5B). However, the term “predetermined” does not mean “fixed”, but may mean that the noise level is derived every time because the noise level may vary depending on the reception environment.

  The first and second correction methods are performed when the reception gain control unit 15 is configured using, for example, a center clipper or AGC.

  The third correction method is performed when the reception gain control unit 15 is configured using, for example, a voice switch.

  In the case of correcting with the first correction method, the section detection correction unit 17 speaks a control signal indicating that the changed threshold value obtained by multiplying the long-term average AvL of the noise level by the attenuation amount (gain) of the reception is used. Output to the speed converter 20.

  When correction is performed by the second correction method, the section detection correction unit 17 outputs a control signal indicating that the noise level included in the reception signal before attenuation by the reception gain control unit 15 is used as a threshold value to the speech speed conversion unit 20. To do. As described above, this control signal includes an instruction for amplification so that the signal level after attenuation becomes the noise level before attenuation according to the attenuation amount of the reception signal by the reception gain control unit 15. .

  In the case of the third correction method, the section detection correction unit 17 transmits a control signal for instructing the use of a predetermined threshold (that is, the long-term average AvL of the reception signal shown in FIG. 5B) to the speech speed conversion unit. 20 is output.

  The speech speed conversion unit 20 has a function of converting the original voice into a voice similar to that when the speaker slowly speaks by expanding the voice signal on the receiving side in the time direction. When speech speed conversion is performed, if the operation is continued at a constant conversion rate, a delay occurs indefinitely with respect to real time. In this case, there is a possibility that the conversation cannot be established on a telephone that performs a real-time call. In the real-time speech speed conversion process, when a speech section of an incoming call is detected and the speech speed of the speech section is reduced, the non-speech section is shortened to recover the delay. Thereby, there is little delay with real time and the effect of reproduction | regeneration (slow reproduction | regeneration) by speech speed conversion is acquired.

  Therefore, the speech speed conversion unit 20 can prevent delay by extending the voice section and shortening the non-voice section. For example, if you want to play the voice of "Hello Hello", extended to be long the voice of the "Hello", "Hello", it is between "Hello" and "Hello", reducing the non-speech section.

  The slow talk button 16 is a switch that can be pressed by the user to switch on / off of the speech speed conversion unit 20. The slow talk button 16 has an indicator 16z that is turned on when the speech speed conversion unit 20 is turned on and turned off when turned off, and is repeatedly turned on and off each time it is pressed. The indicator 16z is provided to indicate the status of the telephone 10.

  FIG. 3 is a block diagram illustrating an example of the internal configuration of the speech speed conversion unit 20 in detail. The speech speed conversion unit 20 includes a speech section detection unit 21, a speech conversion unit 22, a non-speech conversion unit 23, and a signal memory 24.

  The speech section detection unit 21 as an example of the section detection unit includes a noise level estimation unit 21z and a signal amplification unit 21x, and detects a speech section included in the input signal on the reception side.

  The noise level estimation unit 21z estimates the level of the noise signal around the user (that is, the noise level) in the non-voice section included in the input signal on the receiver side.

  The signal amplification unit 21x is used in the second correction method described above, and amplifies the noise signal included in the input signal according to the attenuation amount of the received signal in accordance with the control signal from the section detection correction unit 17.

  The voice conversion unit 22 extends the voice section and delays the voice signal. The voice conversion unit 22 includes a vowel / consonance determination unit 22z and a delay addition unit 22y.

  The vowel / consonant determination unit 22z determines a vowel and a consonant included in the audio signal.

  The delay adding unit 22y delays the vowel sound signal (see FIG. 9A) and does not delay the consonant sound signal (see FIG. 9B).

  The non-speech conversion unit 23 shortens (compresses) the non-speech section.

  The signal memory 24 temporarily stores the input audio signal (input signal), and also stores the audio signal (output signal) output from the audio conversion unit 22 and the non-audio conversion unit 23 temporarily. It is a buffer. The signal memory 24 is a small-capacity memory that is defined by the amount of voice data to be converted into speech speed.

  4A and 4B are timing charts showing audio signals before and after speech speed conversion. FIG. 4A shows an input voice signal (input signal) on the receiving side. FIG. 4B shows an audio signal (output signal) after speech speed conversion. The input signal is a signal before the speech speed conversion, and the speech speed is 100%. On the other hand, the output signal is a signal after the speech speed conversion, and the speech speed is 65%.

  As a result of the speech speed conversion, the voice on the receiving side is similar to that when the speaker speaks quickly and is easy to hear. Therefore, it becomes easy for elderly people, hearing-impaired persons, and the like to hear the voice on the receiving side.

  5A, 5B, and 5C are diagrams for explaining a method of detecting a speech section. FIG. 5A is a timing chart showing the detection result of the voice section. FIG. 5B is a timing chart showing the time average value of the signal waveform. FIG. 5C is a timing chart of the audio waveform.

  When the voice on the receiving side shown in FIG. 5C is input, the voice section detection unit 21 calculates and derives the long-time average AvL and the short-time average AvS of this voice signal. In the long-term average AvL, a sufficiently long time, for example, 3 minutes, 5 minutes, or the like is set to include the time when the speaker is speaking and the time when the speaker is silent. On the other hand, in the short-time average AvS, a short time, for example, 3 seconds, 5 seconds, or the like is set so as to capture the volume (volume) of the voice spoken by the speaker.

  As shown in FIG. 5B, the speech section detection unit 21 regards the long-time average AvL of the speech signal as a noise level and sets it as a threshold for determining the speech section. Further, the voice section detection unit 21 regards the short time average AvS of the voice signal as the voice level. The value of the short-term average AvS is a large value and fluctuates when the speaker is speaking, compared with the value of the long-term average AvL.

  As shown in FIG. 5A, the voice section detection unit 21 detects a section where the voice level is higher than the noise level as a voice section, and detects a section where the voice level is equal to or lower than the noise level as a non-voice section. In the timing chart shown in FIG. 5A, a voice interval is represented by a value 1, and a non-voice segment is represented by a value 0.

  The operation of the telephone 10 having the above-described configuration will be described.

  FIG. 6 is a flowchart illustrating in detail an example of a call operation procedure according to the first embodiment. When the telephone 10 is connected to the telephone of the other party via the public switched telephone network (public analog line), the telephone call is started.

  When the user presses the slow talk button 16 during a call, that is, when voice is transmitted and received in real time, the speech speed conversion unit 20 starts speech speed conversion (slow talk) (S1). The start of the slow talk is notified to the user by the indicator 16z of the slow talk button 16 being lit. The telephone 10 can be set to always perform slow talk in advance. In this case, the slow talk button can be omitted. Further, in this case, a slow talk release button that can be pressed when returning to normal talk may be provided.

  The reception gain control unit 15 determines whether the reception is being attenuated (S2). When the reception is being attenuated, the section detection correction unit 17 inputs the attenuation amount (gain) of the reception obtained from the reception gain control unit 15 and starts correcting the detection of the voice section based on the attenuation amount of the reception. (S3).

  The section detection correction unit 17 corrects the voice section detection by any one or combination of the three correction methods described above (S4).

  In the first correction method, the section detection correction unit 17 outputs a control signal to the noise level estimation unit 21z so as to lower the threshold for detecting the voice section by an amount corresponding to the attenuation of the received call (S4A). .

  In the second correction method, the section detection correction unit 17 amplifies the received signal by the amount corresponding to the attenuation amount by the signal amplification unit 21x, and estimates the noise level from the amplified signal to the noise level estimation unit 21z. In this manner, a control signal is output (S4B).

  In the third correction method, the section detection correction unit 17 outputs a control signal to the noise level estimation unit 21z so as not to estimate the noise level during the attenuation of the received call (S4C). In this case, the noise level estimation unit 21z does not estimate the noise level, and a predetermined threshold is used as a threshold compared with the received signal.

  As described above, the first and second correction methods are effective when the reception gain control unit 15 is configured by a center clipper or AGC. The third correction method is effective when the reception gain control unit 15 is configured by a voice switch.

  The voice section detection unit 21 detects the voice section by comparing the received signal with a threshold value (S5). As described above, this voice section is detected when the long-term average AvL of the voice shown in FIG. 5B is larger than the short-term average AvS. The voice section detection unit 21 determines whether or not the receiving side is a voice section (S6).

  In the case of a voice section, the voice conversion unit 22 performs playback (slow playback) by extending the voice section (S7). On the other hand, if it is a non-speech segment, the non-speech conversion unit 23 compresses the non-speech segment (S8). The processing in steps S7 and S8 is a known technique as described above.

  After the processes in steps S7 and S8, the speech speed conversion unit 20 determines whether or not the speech speed conversion (slow talk) has ended (S9). The slow talk is ended when the slow talk button 16 is pressed again. When the slow talk release button is provided, the slow talk may be ended by pressing the slow talk release button or by terminating the call by on-hook.

  If the slow talk is not finished, the speech speed conversion unit 20 returns to step S2 and repeats the same processing. On the other hand, when the slow talk ends, the speech speed conversion unit 20 ends the operation. When the slow talk is finished, the slow talk button 16 is turned off.

  The received voice when a voice switch is used as the reception gain control unit 15 will be described. FIGS. 7A and 7B are timing charts showing changes in the received voice before and after the voice switch is turned on. FIG. 7A shows the signal waveform of the received voice when the voice switch is off. In the transmission after the voice section of reception, noise due to line echo is generated.

  FIG. 7B shows the signal waveform of the received voice after the voice switch is turned on. As described above, the voice switch determines whether or not there is a transmission, performs transmission / reception switching, and attenuates the reception signal when there is a transmission. When the voice switch is on, the voice switch is turned on and the noise signal on the receiving side is reduced during a period when there is a transmission after reception (period tb in FIG. 7). The voice switch is turned off in the period from the end to the start of the reception, that is, the period in which neither transmission nor reception is performed (periods ta and tc in FIG. 7). When the voice switch is turned off, the noise signal fluctuates greatly, and the voice section detection unit 21 erroneously detects the noise as a voice on the receiving side when the noise signal suddenly increases (rises). As a result, the voice segment detection unit 21 detects a voice segment for a period longer than the actual voice segment of the incoming call.

  The section detection correction unit 17 in the present embodiment corrects the voice section detection so that the periods ta and tc are not included in the voice section by the above-described first, second, or third correction method, and is accurate. A voice section (period td in FIG. 7) is obtained.

  Next, a case where speech speed conversion is performed by distinguishing vowels and consonants from speech including vowels and consonants will be described. FIGS. 8A and 8B are diagrams for explaining a method of speech speed conversion for speech including vowels and consonants. Usually, many voices are composed of vowels and consonants. For example, the case where the voice of “Asa” is spoken slowly is shown as an example. In order to make it easy to distinguish the voice of “ASA” from vowels and consonants, it is expressed as “ASA”, “A” is a vowel, “S” is a consonant, and “A” is a vowel.

  As shown in FIG. 8A, when speaking slowly as compared to normal speaking, the consonant “S” does not increase so much and the vowel “A” tends to extend longer. Such a voice is a natural voice. On the other hand, as shown in FIG. 8B, when the speech speed conversion unit 20 uniformly expands the voice without distinguishing between consonants and vowels, a sense of incongruity different from that of the real voice occurs.

  In the present embodiment, the speech speed conversion unit 20 calculates the degree of vowel-likeness and consonant-likeness of a speech section using an AMDF (Average Magnitude Deference function) value, and the speech speed of the vowel section is slower than that of the consonant section. To convert the speaking speed. The AMDF value is calculated by the speech speed conversion unit 20 in order to obtain the basic period (repetition period) of speech. In detecting the basic period of speech, the degree of correlation (autocorrelation value) between the waveform and the waveform shifted in time is obtained, and the interval (pitch) at which the correlation is strongest is obtained. For the calculation of autocorrelation, there is a method of using integration or subtraction of signals. In this embodiment, the AMDF value is obtained using subtraction (difference) with a relatively small subtraction amount.

  FIGS. 9A and 9B are timing charts showing temporal changes in AMDF values. The vertical axis represents the AMDF value, and the horizontal axis represents the time difference (Δt). FIG. 9A shows the time change of the AMDF value in the vowel section. In the vowel section, the periodicity of the speech signal is strong as shown in FIG. 9A, and in the consonant section, the periodicity of the speech signal is weak as shown in FIG. 9B. The stronger the periodicity of the signal, the smaller the minimum value of the ADMF value appears as a peak, but the maximum value of the ADMF value does not change.

  Therefore, the vowel-likeness is represented by the variable X as shown in the equation (1). The speech speed conversion unit 20 determines that the interval is more likely to be a vowel as the variable X is smaller and closer to the value 0.

  X = AMDF value minimum value mB / AMDF value maximum value mA (1)

  The speech speed conversion unit 20 performs the speech speed conversion so that the variable X becomes smaller and the section becomes a natural voice close to the real voice with a slower voice. Here, the speech speed conversion unit 20 is continuously changed so that the delay amount increases as the value of the variable X decreases. However, by setting the threshold value Th1 and comparing the variable X and the threshold value Th1, A vowel and a consonant may be distinguished. That is, when the variable X <Th1, the speech speed conversion unit 20 determines that it is a vowel, increases the delay amount, and when the variable X ≧ Th1, determines that it is a consonant and sets the delay amount to the value 0. Or you may make it small.

  Further, although the ratio of the minimum value mB of the AMDF value and the maximum value mA of the AMDF value is used as the variable X, the absolute value of the difference between the minimum value mB of the AMDF value and the maximum value mA of the AMDF value may be used. Good. In this case, the vowel and the consonant may be distinguished by comparing the variable X represented by the absolute value of the difference with the threshold Th2. That is, when the variable X> Th2, the speech speed conversion unit 20 determines that it is a vowel, increases the delay amount, and when the variable X ≦ Th2, determines that it is a consonant and sets the delay amount to the value 0. Or you may make it small.

  In this way, speech speed conversion can be performed on speech including vowels and consonants so that speech is close to real voice. Also, by not performing speech speed conversion on consonants, speech speed conversion can be realized with a small amount of computation.

  Note that the voice is not limited to a combination of vowels and consonants, but may be a voice of only consonants, for example, “n”. In the present embodiment, the case where the voice is spoken in Japanese is shown. However, the present invention can be similarly applied to the case where the voice is spoken in a foreign language such as English or German.

  As described above, the telephone 10A (telephone apparatus) in the first embodiment makes a call with the telephone 10B (other telephone apparatus) via the telephone exchange network 7. The microphone 11 collects the user's transmitted voice. The noise level estimation unit 21z estimates the level of a noise signal around the user. The voice section detector 21 detects a voice section included in the transmission signal from the telephone 10B. The reception gain control unit 15 attenuates the reception signal including the line echo component of the transmission signal based on the telephone switching network 7 while the user's transmission voice is collected by the microphone 11. The section detection correction unit 17 (threshold correction unit) is based on the attenuation amount of the received signal by the reception gain control unit 15 and the level of the noise signal estimated by the noise level estimation unit 21z. The threshold used for section detection is corrected. The voice conversion unit 22 (speech speed conversion unit) converts the voice included in the transmission signal from the telephone set 10B based on the threshold value corrected by the section detection correction unit 17 and outputs the voice from the speaker 12.

  As a result, it is possible to prevent erroneous detection of the voice section and perform speech speed conversion that is easy for the user to hear. As a result of the speech speed conversion, the voice on the receiver side is similar to that when the speaker speaks slowly and easily, even when speaking quickly. Therefore, it becomes easy for elderly people, hearing-impaired persons, and the like to hear the voice on the receiving side.

  The reception gain control unit 15 is a center clipper. The section detection correction unit 17 outputs a signal to the noise level estimation unit 21z so as to lower the threshold by the amount of attenuation of the reception signal by the reception gain control unit 15. The voice section detection unit 21 detects a period in which the received signal exceeds the corrected threshold as a voice section.

  Thereby, when the center clipper is used, the threshold value can be matched to the level of the attenuated received signal. Therefore, the detection of the voice section becomes accurate.

  The reception gain control unit 15 is an automatic gain controller (AGC). The section detection correction unit 17 outputs a signal to the noise level estimation unit 21z so as to lower the threshold by the amount of attenuation of the reception signal by the reception gain control unit 15. The voice section detection unit 21 detects a period in which the received signal exceeds the corrected threshold as a voice section.

  Thereby, when the automatic gain controller is used, the threshold value can be matched with the level of the attenuated received signal. Therefore, the detection of the voice section becomes accurate.

  The reception gain control unit 15 is a voice switch. While the reception signal is attenuated by the reception gain control unit 15, the signal amplification unit 21x amplifies the reception signal by the attenuation, and the noise level estimation unit 21z estimates the level of the amplified noise signal.

  Thereby, when the voice switch is used, the level of the attenuated received signal can be adjusted to the threshold value. Therefore, the detection of the voice section becomes accurate.

  The reception gain control unit 15 is a voice switch. The noise level estimation unit 21z stops the estimation of the level of the noise signal while the reception gain control unit 15 attenuates the reception signal. The voice section detector 21 detects a period in which the received signal exceeds a predetermined threshold as a voice section. Thereby, even when the received signal is sporadically attenuated, it is possible to detect a voice section.

  The telephone 10A also has a slow talk button 16 that instructs the start of speech speed conversion. When the start of speech speed conversion is instructed by the slow talk button 16, the speech speed conversion unit 20 converts the speech contained in the transmission signal from the telephone 10B.

  Thereby, the user can start the speech speed conversion at an arbitrary timing, for example, when it is determined that it is difficult to hear the voice of the other party.

  The slow talk button 16 has an indicator 16z. When the start of speech speed conversion is instructed by the slow talk button 16, the indicator 16z lights up. Thereby, the user can easily know that the speech speed conversion is being performed.

(Background and issues leading to the second embodiment)
Speech speed conversion is performed by temporarily storing an input voice signal (input signal) in a signal memory and reading a past signal at a speed slower than that of the input signal.

  When a voice rate conversion device is installed in a telephone with an answering machine function, when the answering machine message stored as an answering machine is played back at a slow speed, if the answering machine message is long, the signal memory ( Free space in the audio buffer is lost. That is, if the speech speed conversion continues for a long time, the delay between input and output increases, and the signal memory has no free space.

  As a result, speech speed conversion cannot be performed until the free space in the signal memory increases, and the effect of speech speed conversion is reduced. In such a case, for example, the user does not convert the speech speed from the middle, but listens to the received speech at the normal playback speed, becomes difficult to hear, and feels uncomfortable with the change in the voice speed of the received speech. In particular, when listening to an answering machine by voice, depending on the requirement, the voice of the answering machine may be long, and it is assumed that the free space of the signal memory is reduced, and the above situation is likely to occur. On the other hand, when the capacity of the signal memory is increased, the cost increases.

  Therefore, in the second embodiment, an example of a telephone device capable of converting the voice speed of a long answering machine even if the signal memory has a small capacity will be described.

(Second Embodiment)
The telephone device of the second embodiment has almost the same configuration as that of the first embodiment. About the same component as 1st Embodiment, the description is abbreviate | omitted by using the same code | symbol.

  FIG. 10 is a diagram showing in detail an example of an internal configuration of a portion related to speech speed conversion in the telephone 10C of the second embodiment. The telephone 10C is a telephone device with a TAM (telephone answering machine) function that can reproduce an answering machine message by a user's operation. The telephone speed conversion unit 20A, the speaker 12, the signal memory 124, and the memory monitoring unit 125 are provided. , A decoder 126, an answering machine voice storage unit 127, and an encoder 128.

  The signal memory 124 temporarily stores an audio signal (input signal) input to the speech speed conversion unit 20A, and temporarily stores an audio signal (output signal) output from the speech speed conversion unit 20A. It is an audio buffer. The signal memory 124 is a small-capacity memory that is defined by the amount of audio data that is converted into speech speed. The signal memory 124 may be provided inside the speech speed conversion unit 20A or may be provided outside as in the present embodiment.

  The speech speed conversion unit 20A has the same configuration as that of the first embodiment except for the signal memory 124. The speaker 12 outputs a recorded voice message that is played back.

  An answering machine voice storage unit 127 stores an answering machine message, and has a relatively large storage area such as a hard disk or a memory card. The encoder 128 compresses the answering machine message stored in the answering machine voice storage unit 127 by a predetermined voice compression method. Examples of the audio compression method include MP3 (MPEG Audio Layer 3), AAC (Advanced Audio Coding), WAV (Windows Media Audio), and the like. The decoder 126 reads out the answering machine message stored in the answering machine voice storage unit 127, and decompresses the answering machine message using the voice decompression method corresponding to the encoder 128.

  The memory monitoring unit 125 monitors the free space of the signal memory 124. When the free space of the signal memory 124 decreases, that is, when the consumption rate (use rate) of the signal memory 124 exceeds the upper limit threshold Sh1, When the instruction to stop reading the answering machine message is issued and the free space of the signal memory 124 increases, that is, when the consumption rate of the signal memory 124 falls below the lower limit threshold Sh2, the voice message reading to the decoder 126 is resumed. To give instructions. Here, the upper threshold value Sh1 is set to a value where the free capacity of the signal memory 124 is close to the predetermined lower limit value. The lower limit threshold Sh2 is set to a value where the free space of the signal memory 124 is close to the predetermined upper limit.

  FIG. 11 is a timing chart showing a change in state in each part of the telephone 10C. In this timing chart, changes with time of reading by the decoder 126, slow sound output, and the consumption rate (usage rate) of the signal memory 124 are shown.

  At timing t0, when the decoder 126 reads the answering machine message and the speech speed conversion unit 20A performs the speech speed conversion to output the slow voice, the consumption rate of the signal memory 124 increases. When the consumption rate of the signal memory 124 exceeds the upper limit threshold Sh1 (for example, a value close to a predetermined upper limit set to about 20% of the maximum capacity) at the timing t1, the memory monitoring unit 125 makes an absence to the decoder 126. Reading of the answering machine message stored in the recorded voice storage unit 127 is stopped. While reading of the answering machine message is stopped, the speech speed conversion unit 20A performs the speech speed conversion and outputs a slow voice.

  Thereafter, when the consumption rate of the signal memory 124 gradually decreases and falls below a lower limit threshold Sh2 (for example, a value close to a predetermined lower limit set to about 80% of the maximum capacity) at timing t2, the memory monitoring unit 125 Then, the decoder 126 is made to resume reading out the answering machine message stored in the recorded voice recording unit 127.

  As a result, the speech speed conversion unit 20A can convert the answering voice over a long period of time even if the signal memory 124 has a small capacity without interrupting the output of the slow voice. Further, by setting an upper limit threshold Sh1 and a lower limit threshold Sh2, and setting the lower limit threshold Sh2 wider than the upper limit threshold Sh1, the speech speed conversion is performed after the free space of the signal memory 124 is sufficiently recovered. Can be resumed, and it can be prevented that the free space is remarkably reduced in the middle of the speech speed conversion. Therefore, the operation of the speech speed conversion is stabilized. In addition, it is possible to reduce the processing load due to frequent stop / restart of the decoder 126.

  FIG. 12 is a flowchart for explaining an example of a call / playback operation procedure according to the second embodiment. The telephone 10C determines whether a call is in progress (S21). When the telephone call is in progress, the telephone set 10C executes the received voice real-time playback mode, and performs the speech speed conversion in the same manner as in the first embodiment (S22). Since the speech speed conversion in step S22 has been described in detail in the first embodiment, a description thereof will be omitted.

  On the other hand, if the telephone call is not being made in step S21, the telephone 10C determines whether or not to reproduce the answering machine message (S23). If the answering machine message is not reproduced, the telephone set 10C returns to the process of step S21.

  When the answering machine message is reproduced, the telephone set 10C shifts to the non-real time reproduction mode of the answering voice, that is, the slow reproduction mode of the answering machine message (S24). In non-real-time playback (playback of an answering machine message) in the second embodiment, slow playback is also performed in the non-speech section. However, as in the first embodiment, slow playback is not performed in the non-speech section. It may be. By not performing slow playback in the non-voice section, the signal memory consumption rate can be quickly reduced. It is also possible to perform speech speed conversion that is easy for the user to hear.

  The decoder 126 reads the answering machine message from the answering voice storage unit 127, decompresses the answering machine message, and stores the decompressed answering machine message in the signal memory 124 (S25).

  The speech speed conversion unit 20A converts the answering machine message in the signal memory 124 to reproduce the speech speed (slow playback) (S26). The memory monitoring unit 125 determines whether the consumption rate of the signal memory 124 has exceeded the upper limit threshold Sh1 and the free space has decreased (S27). If the upper limit threshold Sh1 is not exceeded, the decoder 126 continues to read the answering machine message (S30).

  On the other hand, when the consumption rate of the signal memory 124 exceeds the upper limit threshold Sh1, the memory monitoring unit 125 causes the decoder 126 to stop reading the answering machine message (S28). Then, the memory monitoring unit 125 determines whether or not the consumption rate of the signal memory 124 falls below the lower limit threshold Sh2 and the free space has increased (S29).

  If it is not below the lower limit threshold Sh2, the memory monitoring unit 125 returns to the process of step S28. On the other hand, if the consumption rate of the signal memory 124 falls below the lower limit threshold Sh2, the decoder 126 resumes reading out the answering machine message in step S30.

  Then, the telephone set 10C determines whether or not to finish playing the recorded message (S31). When the reproduction of the recorded message is not finished, the process returns to step S26, and the speech speed conversion unit 20A performs the speech speed conversion and reproduces it. On the other hand, when the reproduction is terminated due to a button operation or when all the answering machine messages in the answering voice storage unit 127 have been reproduced, the telephone 10C terminates this operation.

  As described above, the telephone 10 </ b> C (telephone apparatus) in the second embodiment makes a call with the telephone 10 </ b> B (other telephone apparatus) via the telephone exchange network 7. The encoder 128 (compression unit) compresses the call voice transmitted from the telephone 10B. An answering machine voice storage unit 127 (an answering machine message storage unit) stores the call voice compressed by the encoder 128 as an answering machine message. The decoder 126 (decompression unit) decompresses the call voice stored in the recorded voice recording unit 127. The signal memory 124 (voice buffer) temporarily holds the call voice expanded by the decoder 126. The speech speed conversion unit 20 converts the speech speed read from the signal memory 124 and outputs it from the speaker 12. When the memory monitoring unit 125 (monitoring unit) determines that the free capacity of the signal memory 124 has approached the predetermined lower limit, the memory monitoring unit 125 (monitoring unit) temporarily stops the expansion of the call voice in the decoder 126.

  Thereby, even if the signal memory 124 has a small capacity, it is possible to perform the speech speed conversion for the voice of the long absence recording.

  When the memory monitoring unit 125 determines that the free space of the signal memory 124 has approached the predetermined upper limit value, the voice expansion in the decoder 126 is resumed. As described above, when the free space of the voice buffer increases, the voice speed conversion can be performed without interruption by resuming the expansion of the call voice.

  The predetermined lower limit value is about 20% of the maximum capacity of the signal memory 124. By temporarily stopping the expansion of the call voice at about 20% of the maximum capacity, a margin can be provided in the voice buffer, and the operation of the speech speed conversion is stabilized.

  The telephone 10C has a slow talk button 16 for instructing start of speech speed conversion. When the start of speech speed conversion is instructed by the slow talk button 16, the speech speed conversion unit 20A converts the speech contained in the transmission signal from the telephone 10B.

  Thereby, the user can start the speech speed conversion at an arbitrary timing, for example, when it is determined that it is difficult to hear the voice of the other party.

  The slow talk button 16 has an indicator 16z. When the start of speech speed conversion is instructed by the slow talk button 16, the indicator 16z lights up. Thereby, the user can easily know that the speech speed conversion is being performed.

  In the second embodiment as well, as in the first embodiment, speech speed conversion can be performed on speech including vowels and consonants by distinguishing vowels and consonants.

  While various embodiments have been described above with reference to the drawings, it goes without saying that the present invention is not limited to such examples. It will be apparent to those skilled in the art that various changes and modifications can be made within the scope of the claims, and these are naturally within the technical scope of the present invention. Understood.

  For example, in the above-described embodiment, the case where the telephone device is a fixed telephone is shown, but a mobile telephone that performs a telephone call wirelessly may be used.

  INDUSTRIAL APPLICABILITY The present invention is useful as a telephone device that can suppress erroneous detection of a voice section included in a received signal and can efficiently perform speech speed conversion that is easy for a user to hear.

5A, 5B exchange 10, 10A, 10B, 10C Telephone 11 Microphone 12 Speaker 15 Receive gain control unit 16 Slow talk button 17 Section detection correction unit 20, 20A Speech speed conversion unit 21 Voice section detection unit 21z Noise level estimation unit 22 Voice Conversion unit 21x Signal amplification unit 22y Delay addition unit 22z Vowel / consonation determination unit 23 Non-voice conversion unit 24, 124 Signal memory 30, 30A, 30B 2-wire 4-wire conversion circuit 50 Public analog line 125 Memory monitoring unit 126 Decoder 127 Voice mail recording Voice accumulation unit 128 Encoder AvL Long-time average AvS Short-time average Sh1, Sh2 Threshold

Claims (7)

A telephone device for making a call with another telephone device via a telephone exchange network,
A microphone that picks up the user's transmitted voice;
A noise level estimator for estimating a level of a noise signal around the user;
A section detector for detecting a section of voice included in the transmission signal from the other telephone device;
A reception gain control unit for attenuating a reception signal including a line echo component of the transmission voice based on the telephone switching network while the user's transmission voice is picked up by the microphone;
Threshold correction for correcting a threshold value used for detecting the speech section in the section detection section based on the attenuation amount of the received signal by the reception gain control section and the level of the noise signal estimated by the noise level estimation section. And
A speech rate conversion unit that converts the speech rate included in the transmission signal from the other telephone device based on the threshold value corrected by the threshold value correction unit and outputs it from a speaker;
Telephone device. The telephone device according to claim 1,
The reception gain control unit is a center clipper,
The threshold correction unit lowers the threshold according to the attenuation of the received signal by the center clipper,
The section detection unit detects a period in which the received signal exceeds a threshold corrected by the threshold correction unit as the voice section.
Telephone device. The telephone device according to claim 1,
The reception gain control unit is an automatic gain controller,
The threshold correction unit lowers the threshold according to the attenuation of the received signal by the automatic gain controller,
The section detection unit detects a period in which the received signal exceeds a threshold corrected by the threshold correction unit as the voice section.
Telephone device. The telephone device according to claim 1,
The reception gain control unit is a voice switch,
The noise level estimation unit estimates the level of the noise signal by amplifying the reception signal according to the attenuation while the reception signal is attenuated by the voice switch.
Telephone device. The telephone device according to claim 1,
The reception gain control unit is a voice switch,
The noise level estimation unit stops the estimation of the level of the noise signal while the received signal is attenuated by the voice switch,
The section detection unit detects a period in which the received signal exceeds the predetermined threshold as the voice section.
Telephone device. The telephone device according to any one of claims 1 to 5,
A button for instructing the start of speech rate conversion,
The speech rate conversion unit, when instructed to start speech rate conversion by the button, converts speech rate included in a transmission signal from the other telephone device,
Telephone device. The telephone device according to claim 6,
An indicator showing the status of the telephone device;
When the start of speech speed conversion is instructed by the button, the indicator lights.
Telephone device.

Images