JP2007219188A - Consonant processing device, speech information transmission device, and consonant processing method - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide an inexpensive consonant processing device, an inexpensive speech information transmission device, and a consonant processing method which are capable of transmitting speech information almost in real time, in which signal processing is easy, and in which even the elderly or the people with hearing difficulty can distinguish a consonant even in the environment full of noise. <P>SOLUTION: The device comprises: a frame division part 1 for extracting a frame signal by each of a plurality of time frames from an inputted speech signal; a power calculation part 2 for calculating an average power or a sound pressure level for each frame signal; a comparison part 3 for mutually comparing the average powers and sound pressure levels between the frame signals; a consonant determination part 4 for determining whether the speech signal is a consonant or not based on the comparison result by the comparison part 3; and an amplification part that amplifies the amplification object point or the amplification object width of the speech signal when the consonant determination part 4 has determined that the speech signal is a consonant, but does not amplify it when the consonant determination part 4 has determined that the speech signal is not a consonant or a syllable end. <P>COPYRIGHT: (C)2007,JPO&INPIT

Description

  The present invention provides a consonant processing device that can transmit audio information in real time, has simple signal processing, and is easy to hear the end points of consonants or syllables even in a noisy environment or an elderly person or a hearing impaired person, and can be manufactured at low cost. The present invention relates to a voice information transmission device and a consonant processing method.

  Elderly people with reduced hearing ability and hearing-impaired people are inevitably less able to hear words than ordinary people. For such elderly people and hearing-impaired persons, the contents of various voices in public spaces with a lot of noise, such as on-site broadcasts such as airfields, guidance broadcasts in trains, buses, vending machines and ATM guidance voices, are accurate. It is difficult to hear. In particular, the voice of evacuation guidance in an emergency can lead to a major accident if the contents are not understood.

  Wireless broadcasting systems have been developed that carry receivers and transmit sound via radio or infrared communication media for the elderly and hearing impaired people who have weakened their ability to hear such words. . In such a system, the speaker speaks to a microphone placed near the mouth in a quiet place with little noise, and the clear sound is sent directly to the user via the communication medium. Therefore, the user can always listen to clear sound in any noise environment.

  In addition, hearing aids and other hearing aids are equipped with a noise reduction function that suppresses ambient noise, a consonant enhancement function that emphasizes only consonants in speech as described below, etc. Has been proposed. In particular, consonant emphasis is a technology developed by focusing on the fact that it is difficult for the elderly and the hearing impaired to hear the consonant part because the consonant amplitude of speech is smaller than that of vowels.

  However, such a phenomenon is not limited to that resulting from a decrease in hearing. Even for humans whose hearing ability has not deteriorated, announcement broadcasting devices such as private broadcasting and guidance broadcasting, mobile phones, and other voice information transmission devices cannot hear sound due to noise in a noisy environment. On the other hand, it is only necessary to increase the volume by installing speakers or earphones with high output, but there is a possibility that the permissible limit of hearing may be exceeded, and there is a limit to the size of the device. Distortion etc. become large.

  For this reason, several methods of speech enhancement have been proposed for the purpose of making speech easy to hear. This voice enhancement is to amplify the spectrum amplitude in a predetermined band of the frequency spectrum of the voice and improve the intelligibility of the voice as seen from the listening side. This amplifies the predetermined frequency band power and amplifies it. It is to attenuate the spectrum amplitude of a band different from the band. (See Patent Document 1).

  Here, the mechanism of sound generation will be explained. When the vocal cords vibrate, a voice wave is generated in the vocal tract from the vocal cords to the lips, and the voice wave generated in the vocal tract is voiced through the lips and the tongue. Radiates as. That is, when the vocal cords vibrate at a constant cycle (pitch cycle), the vocal tract resonates air from the lungs with a shape (eg, thickness) such as a throat to generate vowels. At this time, by changing the shape of the throat or the like, the vowels such as “A”, “I”, “U” are tuned and radiated as voice waves. On the lips, tongue, and the like, plosive sounds, friction sounds, nasal sounds, and other sounds are generated in the mouth as consonants. Normally, consonants and vowels are combined and emitted into the space.

  By the way, Japanese is a CV (Consonant-Vowel) type language. For example, the pronunciation of “ka” in Japanese is composed of a consonant (C) “ku” and a vowel (V) “a”. Consonants and vowels are emitted together. For example, English is also a CVC (Consonant-Vowel-Consonant) type language, and it is known that consonants, vowels, and consonants are often arranged in this order.

  Therefore, in the case of Japanese, the pronunciation of the other lines except “n” and “tsu” is pronounced with such a combination of consonant (C) and vowel (V). In each case, the sound from the vocal cords is first disturbed using the lips and tongue, etc., and the consonant is tuned, and then the vowel is uttered with a strong sound without disturbing the sound from the vocal cords.

  For this reason, the consonant has a smaller voice amplitude than the vowel, and the voice of the voice information transmission device cannot be heard in the surrounding environment due to noise. For this reason, for example, a voice information transmission device such as a hearing aid that can clearly hear the voice has been proposed (see Patent Document 2).

  The audio information transmission apparatus of Patent Document 2 includes a microphone to which audio is input from the outside, an audio signal processing unit that generates a consonant clarification signal based on the input audio signal, and a carrier signal generation unit that generates a carrier signal And an amplitude modulation unit that modulates the amplitude of the carrier wave signal based on the consonant clarification signal, and a vibrator that transmits mechanical vibration based on the amplitude-modulated output signal. The consonant part included in the audio signal is extracted by the unit, and the consonant part consonant part extracted by the iterative processing unit is repeated a plurality of times and added to the audio signal to generate a consonant clarification signal.

  The consonant part in the audio signal has a VOT (Voice Onset Time) of about several tens of ms between the vowel part. This VOT is the time from the burst of a consonant until the vocal cords vibrate, and is in a state close to silence. Therefore, the amplitude is smaller than the rise of the consonant part and the vowel, an appropriate reference value is set, and a region where the amplitude is below the reference value continues for a predetermined time (for example, about 10 ms) is determined as VOT. Thus, the end of the consonant part can be identified by distinguishing from the remaining part of the consonant part and the vowel part.

  Similarly, a silent section of several tens of ms or more usually exists between the vowel part and the subsequent consonant part. Therefore, the beginning of the next consonant part is specified by discriminating this silent section in the same manner as the detection of VOT.

  Similarly to Patent Document 2, in Patent Document 1 described above, speech that improves the clarity of received speech from a mobile phone or the like and suppresses deterioration of speech quality and increase in noise when the input speech includes noise. Emphasis devices have been proposed.

  The speech enhancement apparatus of Patent Literature 1 estimates a speech quality of an input speech signal and outputs a speech quality estimation value (estimated S / N ratio), and a speech quality estimation output by the speech quality estimation unit. A voice enhancement processing unit that adjusts the vocal tract characteristics of the input voice signal (formant amplification, anti-formant attenuation) and emphasizes the residual signal of the input voice signal (pitch enhancement) based on the value Is. The residual signal is a sound source signal separated by removing a linearly predictable portion from a speech wave, and the sound source pitch period can be obtained by calculating the autocorrelation thereof.

JP 2005-331783 A JP 2005-287600 A

  As described above, there is a problem that the consonant is weaker than the vowel, and the voice of the voice information transmission device cannot be heard due to noise depending on the surrounding environment.

  The above-described wireless broadcasting system can only be used in a limited public space equipped with such a system, and the user must carry a receiver. Moreover, since this broadcasting system is large and expensive, it is difficult to install it in any public space, and it is difficult for all users to carry the receiver, so that there is a problem that it is difficult to spread.

  There are also problems with hearing aids and the like. First, it is useless if the user is not wearing it, and the input to the hearing aid is a sound in which voice and ambient noise are mixed. Therefore, it is necessary to suppress only the noise or to emphasize only the consonant from the mixture of both sounds by the noise suppression mechanism and the consonant enhancement mechanism. However, they may not work well depending on the type of noise, and consonant enhancement is difficult even in silence. For this reason, in the past, detection accuracy has been improved by detecting consonants using multiple cues in parallel, such as amplitude envelopes, silence intervals associated with bursts of silent closed consonants, and other frequency domain information. . However, the processing for this is complicated, and it has been an obstacle to performing speech information transmission by emphasizing consonants in real time (real time) or a time close to this (semi-real time).

  The speech enhancement device of Patent Literature 1 estimates a S / N ratio of an input speech signal, and performs a process of giving a positive gain to formant power centered on the formant frequency based on the S / N ratio. In addition, a process of giving a negative gain to the antiformant power centered on the antiformant frequency is performed, and the pitch is emphasized to make it easy to hear. However, the processing is complicated, time consuming, expensive, and there are many problems in performing speech information transmission processing in a time close to real time. The speech enhancement apparatus of Patent Document 1 does not take advantage of the characteristic that Japanese speech is a combination of consonants and vowels, and the consonants are weaker than vowels.

  On the other hand, the speech information transmission apparatus of Patent Document 2 extracts the consonant part included in the speech signal, and the extracted consonant part consonant part is repeated a plurality of times and added to the speech signal to add the consonant clarification signal. Is generated. For this reason, although it has improved voice discrimination, the consonant part is repeated in each sound, and there is a delay in the consonant part of each sound. There wasn't. In order to discriminate between VOT and silent section, information on this VOT and silent section is necessary. For this reason, it is necessary to store the audio signal once and process it, and the processing is complicated and takes time. It is the same as Reference 1.

  As described above, the conventional technique uses a plurality of cues in parallel to detect consonants, and therefore the consonant enhancement processing is very complicated, and voice information can be transmitted in real time or in a time close to this. It was not a thing. In addition, these techniques need to store sound that has been subjected to enhancement processing in advance, and are difficult to use in a voice information transmission device that requires flexibility.

  Therefore, the present invention is a consonant that can transmit voice information in a time close to real time, is simple in signal processing, is easy to hear the end points of consonants and syllables even in noisy environments, elderly people and hearing impaired people, and can be manufactured at low cost. An object is to provide a processing device and a voice information transmission device.

  In addition, the present invention provides a consonant processing method that can transmit voice information in a time close to real time, is simple in signal processing, and is easy to hear the end points of consonants and syllables even in a noisy environment, even in elderly people or hearing impaired people. The purpose is to do.

  A consonant processing device according to the present invention includes a frame dividing unit that extracts a frame signal from a plurality of time frames from an input audio signal, a power calculation unit that calculates an average power for each of the frame signals, and a frame signal. A comparison unit that compares average power with each other, a consonant determination unit that determines whether the amplification target point or amplification target width of the audio signal is an end point of a consonant or syllable based on the comparison result of the comparison unit, and a consonant determination unit Amplifying the sound signal amplification target point or amplification target width when it is determined as a consonant or syllable end point, and an amplification unit that does not amplify when it is determined not to be a consonant or syllable end point And

  According to the consonant processing device, the speech information transmission device, and the consonant processing method of the present invention, it is possible to perform consonant enhancement only by extracting a plurality of frame signals from a plurality of time frames and calculating and comparing the average power of the frame signals. Therefore, it is not necessary to perform various processes in parallel, voice information can be transmitted in a time close to real time, signal processing is simple, even in situations where noise or voice competes with other acoustic signals. In addition, it becomes easier for hearing-impaired and elderly people to hear the end points of consonants or syllables, thereby reducing the overall strength of the speech without impairing the clarity of the speech and preventing the increase in environmental noise. it can. Further, it is possible to provide a consonant processing device and a voice information transmission device that can be manufactured at low cost.

  According to a first aspect of the present invention, a frame division unit that extracts a frame signal from each of a plurality of time frames from an input audio signal, a power calculation unit that calculates an average power for each of the frame signals, and a frame signal A comparison unit that compares the average power with each other, a consonant determination unit that determines whether the amplification target point or the amplification target width of the audio signal is an end point of a consonant or a syllable based on a comparison result of the comparison unit, and a consonant determination And an amplification unit that amplifies the amplification target point or amplification target width of the audio signal when it is determined to be an end point of a consonant or syllable, and does not amplify when it is determined that it is not an end point of a consonant or syllable. Is a consonant processing device. With this configuration, it is possible to perform consonant enhancement simply by extracting multiple frame signals using multiple time frames, calculating the average power of these frame signals, and comparing them, making it unnecessary to perform various processes in parallel. Audio information can be transmitted within a short time, signal processing is simple, even under noisy conditions, or when the voice is competing with other acoustic signals. This makes it easy to hear, thereby reducing the strength of the entire voice without losing the clarity of the voice, preventing an increase in environmental noise, and providing a consonant processing device that can be manufactured at low cost. .

  According to a second aspect of the present invention, a frame dividing unit that extracts a frame signal from a plurality of time frames from an input audio signal, a power calculation unit that calculates an average power for each of the frame signals, and a frame signal A comparison unit that compares the average power with each other, a consonant determination unit that determines whether the audio signal is an end point of a consonant or a syllable based on a comparison result of the comparison unit, and a consonant determination unit that is an end point of a consonant or a syllable If determined, the amplification target point of the audio signal or the amplification degree of the amplification target width is determined in the amplification direction, and if it is determined that it is not the end point of the consonant or syllable, the amplification degree determination unit determines that the audio signal is not amplified, A consonant processing apparatus comprising: an amplifying unit that amplifies an audio signal in accordance with the amplification degree determined by the amplification degree determining unit. With this configuration, in addition to the effects of the first embodiment, the amplification degree can be adjusted by the amplification degree determination unit, and a consonant processing device that can be easily heard can be provided.

  A third aspect of the present invention is a form subordinate to the first or second form, characterized in that the comparison unit performs comparison by calculating a difference in decibel average power of each frame signal. This is a consonant processing device that simply calculates the difference, so that signal processing can be easily performed, and voice information can be transmitted in a time close to real time.

  A fourth form of the present invention is a form subordinate to the first or second form, wherein the comparison unit compares by calculating the ratio of the average power of each frame signal. Since it is a device and only calculates the ratio, it can easily process signals and transmit voice information in a time close to real time.

  According to a fifth aspect of the present invention, in the consonant processing device according to any one of the first to fourth aspects, the time frame is provided with a time frame having an extraction width from which a consonant can be extracted. Is a consonant processing device characterized in that it is set at the center position of the extraction width of this time frame, VCV type signal processing can be suitably performed, and the configuration can be simplified and effectively amplified.

  According to a sixth aspect of the present invention, in the consonant processing device according to any one of the first to fourth aspects, when two time frames continuous to the time frame are provided, the amplification target point or the amplification target width is 2 It is a consonant processing device characterized in that it is set at the boundary of two time frames, CV type signal processing can be suitably performed, the configuration is simple, and it can be amplified effectively.

  In the consonant processing device according to the third aspect of the present invention, in the consonant processing device according to the third aspect, when the difference in average power displayed in decibels is 0 or less, the amplitude of the audio signal at the amplification target point or the amplification target width is amplified. The consonant processing device is characterized in that it does not amplify when the decibel-displayed difference is larger than 0. In addition to the operational effects of the above-described embodiment, signal processing is further simplified.

  According to an eighth aspect of the present invention, in the consonant processing device of the third aspect, when the ratio between the average powers is 1 or less, the amplitude of the audio signal at the amplification target point or the amplification target width is amplified and the average 5. The consonant processing device according to claim 4, wherein the power is not amplified when the power ratio is greater than 1, and in addition to the function and effect of the above aspect, the signal processing is further simplified.

  According to a ninth aspect of the present invention, in the consonant processing device according to any one of the first, third to sixth aspects, when the amplifying unit determines that the consonant determining unit is an end point of a consonant or a syllable, the sound signal is amplified. Instead of amplifying the target point or the amplification target width, a consonant processing device is characterized in that when it is determined as a consonant or an end point of a syllable, the amplification target point or the amplification target width of the audio signal is suppressed. With this configuration, it is possible to provide a sound that is difficult to hear and to be used for hearing tests and listening training.

  According to a tenth aspect of the present invention, in the consonant processing device according to any one of the second to sixth aspects, when the amplification degree determination unit determines that the consonant determination unit is an end point of a consonant or a syllable, Instead of determining the amplification degree of the point or the amplification target width in the amplification direction, the amplification degree determination unit determines that the consonant determination unit determines that it is an end point of a consonant or syllable, or the amplification target point or amplification target of the audio signal A consonant processing device that performs a determination to reverse the width amplification degree in a suppression direction. With this configuration, the amplification target point or the amplification target width is suppressed, and it is possible to provide a sound that is difficult to hear and to be used for hearing test or listening training.

  According to an eleventh aspect of the present invention, in the consonant processing device according to any one of the first to eighth aspects, a filter unit that allows a predetermined frequency component to pass is provided before an audio signal is input to a frame division unit between average powers. The consonant processing apparatus is characterized by the fact that the clarity of consonant enhancement can be increased.

  A twelfth aspect of the present invention is a form subordinate to any one of the first to eleventh aspects, in which the amplifying unit determines the loudness as a sensory amount of a sound-sensitive deaf person. The consonant processing device according to claim 1, wherein the physical sound pressure is amplified according to a correction characteristic of a supplementary phenomenon that matches a volume of a sound, which is a volume, and is listened to by a normal listener It is possible to amplify the end points of consonants or syllables to the strength that is felt easy.

  According to a thirteenth aspect of the present invention, a consonant processing device subordinate to any one of the first to twelfth embodiments and a consonant-enhanced sound based on a consonant-processed audio signal from the consonant processing device are output. A voice information transmission device including a speaker. With this configuration, it is not necessary to perform various processes in parallel, voice information can be transmitted in a time close to real time, signal processing is simple, and there is a situation where noise or voice competes with other acoustic signals. However, it is easier for deaf and elderly people to hear the end points of consonants or syllables, thereby reducing the overall strength of the speech without compromising the clarity of the speech and preventing an increase in environmental noise. Therefore, an inexpensive audio information transmission device can be provided.

  In the fourteenth aspect of the present invention, a frame signal is extracted for each of a plurality of time frames from an input audio signal, an average power is calculated for each of the frame signals, and the average power is compared between the frame signals. Based on the comparison result, it is determined whether or not the amplification target point or amplification target width of the audio signal is a consonant or syllable end point. In this consonant processing method, the target width is amplified and is not amplified when it is determined that the target width is not a consonant. With this configuration, it is not necessary to perform various processes in parallel, voice information can be transmitted in a time close to real time, signal processing is simple, and there is a situation where noise or voice competes with other acoustic signals. However, the end points of consonants or syllables can be easily heard even by a hearing-impaired person or an elderly person, thereby reducing the strength of the entire voice and preventing an increase in environmental noise.

  According to a fifteenth aspect of the present invention, in the consonant processing method according to the fourteenth aspect, when it is determined that the end point of a consonant or a syllable, a consonant or a consonant A consonant processing method is characterized in that when it is determined as an end point of a syllable, the amplification target point or the amplification target width of the voice signal is conversely suppressed. With this configuration, it is possible to provide a sound that is difficult to hear and to be used for hearing tests and listening training.

  A thirteenth aspect of the present invention is a form subordinate to the twelfth aspect, in which the amplification level is the amount of sound that is a sensory amount of a hearing-impaired deaf person, and the volume of sound that is a sensory amount of a normal hearing person. A consonant processing method characterized by amplifying a physical sound pressure in accordance with a correction characteristic of a supplementary phenomenon that matches the depth, and amplifying the end points of consonants or syllables in a state where it is felt that a normal hearing person can easily hear. it can.

Example 1
Hereinafter, a consonant processing device, a voice information transmission device, and a consonant processing method according to Embodiment 1 of the present invention will be described.

  The consonant enhancement performed by the sound enhancement processing apparatus according to the first embodiment is performed when the syllable has a CV-type language structure as in Japanese, that is, when the frequency of consonant (the vowel (V) immediately follows C is high). In this specification, the term “consonant enhancement” is used to emphasize not only consonants but also the end points of syllables.

  FIG. 1 is a configuration diagram of a consonant processing device and a voice information transmission device equipped with the consonant processing device in Embodiment 1 of the present invention. FIG. 2 is an explanatory diagram of processing of the consonant processing device in Embodiment 1 of the present invention. FIG. 5 is an explanatory diagram of the amplification degree at the time of amplification in Example 1 of the present invention.

  In FIG. 1, reference numeral 1 denotes a frame dividing unit that extracts frame signals in a plurality of time frames as shown in FIG. 2 when an audio signal is input. 1a constitutes the frame dividing unit 1 and includes a first time frame for extracting a frame signal having a width of about 1/3 of the length of the consonant, and 1b includes the first time frame 1a. Is a second time frame having an extraction width that allows extraction of the syllable, and 1c is a third time frame that includes the second time frame 1b and that can extract about 1 to 3 times the syllable length.

  The first time frame 1a, the second time frame 1b, and the third time frame 1c have a function of multiplying a window function such as a rectangular window or a Hamming window. In the first embodiment, a rectangular window is employed.

That is, for the audio signal at the time of t = T, the first time frame 1a has the window function w ₁ (t) = 1 (where T−τ ₁ ≦ t ≦ T + τ ₁ ), w ₁ (t) = 0. The second time frame 1b has a window function w ₂ (t) = 1 (where T−τ ₂ ≦ t ≦ T + τ ₂ ), w ₂ (t) = 0 (other than that) ), And the third time frame 1c is also configured by the window function w ₃ (t) = 1 (where T−τ ₃ ≦ t ≦ T + τ ₃ ) and w ₃ (t) = 0 (otherwise). In either case, the unit is ms.

  Note that the center positions (t = T) of the first time frame 1a, the second time frame 1b, and the third time frame 1c do not have to coincide with each other, but in the waveform of the audio signal shown in FIG. This center position is a point to be amplified of the audio signal in these time frames. This may not only amplify the points but also amplify a predetermined width. In this case, this is referred to as an amplification target width in this specification. This amplification position or amplification target width is preferably set at least at the center position of the second time frame 1b. Note that the second time frame 1b may be provided more quickly than the rear end of the third time frame 1c, and the processing speed may be increased. It is preferable to arrange.

The window functions τ ₁ , τ ₂ , and τ ₃ are empirically determined parameters. In the first embodiment, τ ₁ = about 7.5 ms, τ ₂ = about 25 ms (the length that can extract consonants), τ ₃ is set to about 200 ms (a length capable of extracting about 1 to 4 times the syllable length). In the case of Japanese, the length of a consonant is generally about several tens of ms, and the length of one syllable is about 100 to 400 ms.

Therefore, the audio signal p (t), from the first time frame _{1a y 1 (t) = w} 1 (t) · p (t) is output from the second time frame 1b _y 2 (t ) = W ₂ (t) · p (t), and the frame signal is extracted from the third time frame 1c by y ₃ (t) = w ₃ (t) · p (t). In the case of a digital signal, for example, y ₃ (t) will be described. Between T−τ ₃ ≦ t <T and T <t ≦ τ ₃ + T, N time series values and t = T time series values, respectively. And is calculated with (2N + 1) time series values in total. The same applies to the time series values of y ₁ (t) and y ₂ (t), and the same value as the input time series value of y ₃ (t) is used.

  As described above, the consonant processing apparatus according to the first embodiment performs A / D conversion by the frame division unit 1 and the like, and performs digital processing by the digital circuit or processor including the determination of the amplification degree described later. Can also be used for analog processing. When digital processing is performed by the processor, although not shown, a memory is provided to store programs and set values, which are read and calculated.

Next, in FIG. 1, reference numeral 2 denotes frame signals y ₁ (t), y ₂ (t), and y ₃ (t) extracted in the first time frame 1a, the second time frame 1b, and the third time frame 1c. It is a power calculation part which calculates average power. 2a is a first power calculating unit calculates the average power L ₁ of the average power P ₁ is the square of the amplitude and decibels of y ₁ output from the first time frame 1a _(t). Similarly, 2b denotes a second power calculating unit calculates the average power L ₂ and the average power P ₂ of y ₂ output from the second time frame 1b _(t) and decibels. Furthermore, 2c is a third power calculation unit calculates the average power L ₃ that the average power P ₃ of y ₃ output _(t) and decibels from the third time frame 1c. The average power P _i (i = 1, 2, 3) is expressed by (Equation 1), and 2N + 1 is the total number of time series values. The unit of the average powers L ₁ , L ₂ , L ₃ is dB.

Hereinafter, the difference is described using the average power L _i (i = 1, 2, 3) expressed in decibels, but the ratio of the average power P _i K _ij = P _i / P _j (i, j = 1, 2, 3; i <j). The case where this ratio K _ij is used will be described in Example 4 described later. Further, the same effect can be obtained by calculating the difference P _i -P _j of the average power P _i itself instead of taking the logarithm of the average power P _i and making it L _i (i = 1, 2, 3) in decibels. An effect is obtained. However, the detailed description will be omitted to the description of the decibel display.

Subsequently, 3 is a comparison unit that calculates and compares the difference between the average powers L ₁ , L ₂ , and L ₃ of each frame signal, and 4 indicates whether the audio signal is a consonant based on the comparison result of the comparison unit 3. It is a consonant determination unit that determines whether or not. Further, 5 is a delay unit that delays the audio signal or buffers the data for a time required for the process until amplification, and 6 is a change in the amplification degree of the amplification target point of the audio signal when the consonant determination unit 4 determines that it is a consonant. When it is determined that it is not a consonant, the amplification unit does not change the amplification degree.

  Reference numeral 10 denotes a consonant processing device according to the first embodiment that emphasizes and outputs a consonant when an audio signal is input, and is configured as an integrated circuit for fixed use. Further, 11 is a microphone for inputting sound, 12 is a speaker for outputting sound, and 20 is a sound information transmission device equipped with the consonant processing device 10.

  The audio information transmission device 20 outputs the audio signal emphasized by the consonant by the consonant processing device 10 from the speaker 12, and the above-described wireless broadcasting system, announcement broadcasting device such as private broadcasting and guidance broadcasting, and portable information equipment such as a portable terminal. It can be used for other audio information transmission devices and hearing aids. In the case of the voice information transmission device 20 that does not include the microphone 11, for example, in the case of a vending machine or an ATM guide voice, the sound processing by the consonant processing device 10 may be performed on a prerecorded sound signal. .

  The consonant processing apparatus 10 of the first embodiment is an apparatus effective for a language having a CV type structure in which consonants and vowels continue like Japanese, and the comparison unit 3 uses the structure as follows. The average power of each frame signal is compared with such a reference, and the consonant determination unit 4 determines whether or not it is an end point of a consonant or a syllable.

That is, firstly, when the average power L ₁ of the decibel display is higher than the average power L _{2 of the} decibel display by a predetermined threshold (5 dB in the first embodiment) or more (that is, L ₁ > L ₂ +5), 15 ms (consonant) Therefore, this increase is regarded as an increase in noise. The comparison unit 3 calculates L ₁ -L ₂ and calculates whether it is greater than or less than the threshold value. When it is larger than the threshold, the consonant determination unit 4 determines that the voice signal is noise. If it is less than or equal to the threshold value, it is determined according to the following criteria.

Second, if L ₁ -L ₂ is less than or equal to the threshold (5 dB) and L ₂ <L ₃ , the average power L ₂ during 50 ms (slightly longer than the length of the consonant) in the second time frame 1b but will be lower than the average power L ₃ of 400ms third time frame 1c (length of several syllables), herein it is considered to be the end point of the consonant or syllable.

That is, when the consonant or vowel continues, the end point of the consonant or syllable has a lower average power than the vowel, so if the L ₂ and L ₃ levels are compared and L ₂ is small, the consonant or syllable is displayed in the second time frame 1b. It is estimated that there is an end point. This state is shown in FIG.

In FIG. 2, the average power L ₂ of the frame signal extracted by the window function w ₂ (t) is small, and the average power L ₃ of the frame signal extracted by the window function w ₃ (t) is larger than the average power L _2. . Therefore, it can be estimated that the frame signal extracted in the second time frame 1b is an end point of a consonant or a syllable, and vowels exist before and after, or in the previous part or the subsequent part. At this time, the comparison unit 3 calculates L ₂ −L ₃ , and if L ₂ <L ₃ , the consonant determination unit 4 determines that the frame signal of the second time frame 1b is the end point of the consonant or syllable and performs amplification. .

Accordingly, the amplification unit 6 of the first embodiment does not perform amplification when L ₁ −L ₂ > 5 based on the determination of the consonant determination unit 4 described above, and L ₁ −L ₂ <5. , L ₂ −L ₃ is within a range of 0 to −20 dB, a constant amplification factor λ ₀ , for example, 10 dB is amplified. However, even if L ₁ −L ₂ <5, when the condition of L ₂ −L ₃ <−20 is satisfied, the amplification unit 6 weakens the degree of amplification because it becomes difficult to distinguish from noise. At this time, the consonant determination unit 4 may make this determination. Such an amplification characteristic is illustrated in FIG. Consonant enhancement can be easily performed with a very simple configuration. Note that FIG. 7A is given as an example. If the amplification is suddenly stopped, the amplification level changes discontinuously and the sound becomes uncomfortable, so a reduction method such as a one-dot chain line, or It is preferable to use a smoother reduction method.

  As described above, the consonant processing device, the speech information transmission device, and the consonant processing method according to the first embodiment extract a plurality of frame signals by using a plurality of time frames, and calculate and compare the average power of the frame signals to enhance the consonant. Therefore, there is no need to perform various processes in parallel, voice information can be transmitted in a time close to real time, signal judgment is simple, signal processing is simple, and noise or voice competes with other acoustic signals. This makes it easier to hear consonant sounds even for people who are hard of hearing or for elderly people, thereby reducing the overall strength of the voice without compromising the clarity of the voice and preventing an increase in boundary noise. be able to. Further, it is possible to provide a consonant processing device such as a consonant enhancement processing device and a voice information transmission device that are simple in configuration and can be manufactured at low cost.

(Example 2)
Hereinafter, the consonant processing device, the voice information transmission device, and the consonant processing method according to the second embodiment of the present invention will be described. The consonant processing device and the voice information transmission device according to the second embodiment adjust the amplification degree according to the comparison result of the comparison unit when it is determined as a consonant. The consonant enhancement according to the second embodiment is also particularly suitable for consonant enhancement in a CV type language.

  FIG. 3 is a block diagram of a consonant processing device and a voice information transmission device equipped with the consonant processing device according to the second embodiment of the present invention, FIG. 4 (a) is an explanatory diagram of a supplementary phenomenon, and FIG. 4 (b) is a sound in a quiet environment. FIG. 7B is an explanatory diagram of an amplification factor at the time of amplification in the second embodiment of the present invention. The same reference numerals are used in the second embodiment and the first embodiment, and the processing of the audio signal is the same as that of the first embodiment except that the amplification degree is determined according to the comparison result. Description of is omitted.

  In FIG. 3, 1 is a frame dividing unit, 1a is a first time frame, 1b is a second time frame, and 1c is a third time frame. 2 is a power calculation unit, 2a is a first power calculation unit, 2b is a second power calculation unit, 2c is a third power calculation unit, 3 is a comparison unit, 4 is a consonant determination unit, 5 is a delay unit, Reference numeral 6 denotes an amplification unit. Reference numeral 10 denotes a consonant processing apparatus according to the second embodiment, reference numeral 11 denotes a microphone, reference numeral 12 denotes a speaker, and reference numeral 20 denotes a voice information transmission apparatus according to the second embodiment. These are the same configurations as those in the first embodiment.

The characteristic point of the second embodiment is that the amplification factor λ is adjusted according to the comparison result of the comparison unit 3 while the first example is amplified with a constant amplification factor λ ₀ . In FIG. 3, reference numeral 7 denotes an amplification factor determination unit that determines the value of the amplification factor λ.

When the sound comparing unit 3 determines that the end point of the consonant or syllable is determined, the amplification degree determining unit 7 determines the amplification target point or the amplification target width of the audio signal and determines that it is not the end point of the consonant or syllable. Decide not to amplify the audio signal. If L ₂ <L ₃ in the decibel display, the amplification factor λ is set to a larger value as the level difference between L ₂ and L ₃ is larger when it is determined that the audio signal is not noise.

  Here, the amplification factor λ performed in the second embodiment will be described. The amplification factor λ of the second embodiment employs a characteristic similar to the correction characteristic of the supplementary phenomenon for the sound-sensitive deaf person. This supplementary phenomenon means that the increase in sound volume, which is the amount of sensation, with respect to the increase in the sound pressure of a physical stimulus sound, This is a larger phenomenon.

  For this reason, in the replenishment phenomenon, as shown in FIG. 4A, the louder the sound with a predetermined range, the larger the sound is corrected and the smaller the sound is corrected. As a result, it is possible to correct a weak sound that is compatible with the human auditory system and is difficult to hear into a sound that is easy to hear. FIG. 4B is an experiment of how much difference occurs between a quiet environment and noise that can be subjectively recognized as noise. In the figure, P is the intensity of sound when listening in a quiet environment, and N is the intensity of sound when listening to the same sound in white noise. From this result, N shows the same characteristics as the replenishment phenomenon shown in FIG.

Therefore, the amplification performed in the second embodiment adjusts the amplification degree of the audio signal at the amplification target point as follows. First, when the average power _{L 1} of decibels or more decibels of average power _{L 2} than the predetermined threshold value (Example 2, 5 dB) high (i.e. _{L 1>} L 2 +5) is, 15 ms (consonant length Since the amplitude only increases in a very narrow range of about 1/3 of the above), this increase is regarded as an increase in noise, and L ₁ -L ₂ is calculated to be larger than the threshold value or less Calculate. When it is larger than the threshold, the consonant determination unit 4 determines that the voice signal is noise. If it is less than or equal to the threshold value, it is determined according to the following criteria.

Second, if L ₁ -L ₂ is equal to or less than the threshold (5 dB) and L ₃ -20 <L ₂ <L ₃ , the consonant determination unit 4 determines that the end point of the consonant or syllable is the amplification factor λ Is determined as λ = c · (L ₃ −L ₂ ). Here, c = 0.9. It can be expressed not by the average power difference (L ₃ −L ₂ ) expressed in decibels but by the ratio of the average powers P ₂ and P ₃ , K ₂₃ = P ₂ / P ₃ . At this time, λ becomes λ = (K ₂₃ ^1/2 ) ^d . The meaning of the coefficient c will be described in Example 4 with reference to FIG. At this time, the larger the level difference between L ₃ and L ₂ , or the larger K ₂₃ , the larger the amplification, and the 20 dB difference can be compressed to 2 dB. When L ₂ −L ₃ = −20 dB, the amplification factor λ becomes maximum at 18 dB.

Furthermore, when L ₁ -L ₂ is equal to or less than the threshold value (5 dB) and L ₂ <L ₃ -20, if L ₂ is 20 dB or more lower than L ₃ , the power is small compared with the preceding and following audio signals, and it is impossible. Even if it is amplified, it becomes difficult to distinguish it from noise, so the degree of amplification is gradually reduced. For example, as shown in FIG. 7B, the amplification factor λ is decreased by 4.5 dB every time L ₂ -L ₃ decreases by 10 dB, and the amplification factor λ is set to 0 when L ₂ -L ₃ is −60 dB. However, FIG. 7 (b) is given as an example. If the stepwise reduction is performed, the sound becomes uncomfortable where the amplification degree changes discontinuously. It is preferable to If possible, it is preferable to perform a smoother reduction method that does not cause a sudden change.

The consonant processing device, the speech information transmission device, and the consonant processing method described above are effective as a consonant enhancement processing device, a speech information transmission device equipped with the consonant enhancement processing device, and a consonant enhancement method. As described above, L ₃ −20 < In the case of L ₂ <L ₃ , the amplification level is set to a positive value and the audio signal is amplified in the amplification direction, but conversely, the audio signal can be set to the suppression direction by making the amplification level λ negative. For example, when performing hearing tests for hearing impaired persons, listening training for foreign language learners, etc., continuing to listen to noise over a long period of time results in decreased hearing and discomfort. It becomes a consonant suppression processing apparatus and a consonant suppression method effective for such a hearing test apparatus and a listening training apparatus, and it becomes possible to process speech.

  As described above, the second embodiment is extremely effective not only for CV type languages in which important information is often present at the beginning of syllables, but also for various other languages such as English. Consonant enhancement can be easily performed with a simple configuration. Note that parameters such as the extraction width, bias, and maximum gain of the time frame 1 can be changed according to the sound environment and the purpose of use.

  Since the consonant processing device, the voice information transmitting device, and the consonant processing method of the second embodiment extract the frame signal by each of a plurality of time frames, and calculate and compare the average power of the frame signal, the consonant enhancement can be performed. There is no need to perform various processes in parallel, voice information can be transmitted in near real time, there are few logic judgments, signal processing is simple, and there is a situation where noise or voice competes with other acoustic signals. However, consonant sounds can be easily heard even by a hearing impaired person or an elderly person, thereby reducing the strength of the entire voice and preventing the environment from becoming noisy. Further, it is possible to provide a consonant processing device such as a consonant emphasis processing device and a voice information transmission device that can be widely used for consonant emphasis in many languages and that the amplification degree can be easily adjusted.

  Further, when the amplification degree is set to a negative value by using it as a consonant suppression processing device or the like, the voice can be processed into a voice in which the end point of the consonant or syllable is difficult to hear, and can be used for hearing test, listening training and the like.

(Example 3)
A consonant processing device, a voice information transmission device, and a consonant processing method according to Embodiment 3 of the present invention will be described. The consonant processing device and the voice information transmission device of the third embodiment are also suitable for consonant enhancement of a CV type language.

  FIG. 5 is a configuration diagram of a consonant processing device and a voice information transmission device equipped with the consonant processing device according to the third embodiment of the present invention. FIG. 6 is an explanatory diagram of processing of the consonant processing device according to the third embodiment of the present invention. These are explanatory drawings of the degree of amplification at the time of amplification in Example 3 of the present invention. It should be noted that the same reference numerals in the third and second embodiments have the same configuration, and only the configuration of the time frame is different, and the other points are the same as in the second embodiment. The description of the basic configuration of the apparatus will be omitted to the second embodiment.

  In FIG. 6, 1 is a frame dividing unit, 1a is a first time frame, and 1b is a second time frame. 2 is a power calculation unit, 2a is a first power calculation unit, 2b is a second power calculation unit, 3 is a comparison unit, 4 is a consonant determination unit, 5 is a delay unit, 6 is an amplification unit, and 7 is amplification. It is a degree determination part. Reference numeral 10 denotes a consonant processing apparatus according to the second embodiment, reference numeral 11 denotes a microphone, reference numeral 12 denotes a speaker, and reference numeral 20 denotes a voice information transmission apparatus according to the second embodiment. These are the same configurations as those in the second embodiment.

  The characteristic point of the third embodiment is that the second embodiment extracts 1 to 3 syllables by the third time frame 1c and determines amplification based on the syllables of the front part or the front part or the rear part. On the other hand, the fourth time frame 1d is provided continuously to the second time frame 1b, and the amplification factor λ is adjusted by comparing with the subsequent syllable.

In FIG. 6, reference numeral 1d denotes a fourth time frame, which is provided immediately after the second time frame 1b with an extraction width capable of extracting a consonant. Further, 2d is the third power calculating unit calculates the average power L ₄ that the average power P ₄ was decibels of fourth time frame 1d frame signal y ₄ output from _(t). The average power _{P 4} in equation (1), and i = 4.

The first time frame 1a according to the third embodiment includes a window function w ₁ (t) = 1 (where T−τ ₁ ≦ t ≦ T + τ ₁ ) and w ₁ (t) = 0 (otherwise), The second time frame 1b has a window function w ₂ (t) = 1 (where T−τ ₂ ≦ t ≦ T + τ ₂ ), w ₂ (t) = 0 (otherwise), and the fourth time frame 1d has a window The function w ₄ (t) = 1 (here, T + τ ₂ ≦ t ≦ T + τ ₂ + 2τ ₄ ) and w ₄ (t) = 0 (otherwise). The unit is ms. Although τ ₂ = τ ₄ , τ ₂ and τ ₄ can be different parameters.

These τ ₁ , τ ₂ , and τ ₄ are determined empirically. In the second embodiment, τ ₁ = 7.5 ms and τ ₂ , τ ₄ = 25 ms are set. Accordingly, w ₁ (t) = 1 (here, T−7.5 ≦ t ≦ T + 7.5), w ₁ (t) = 0 (otherwise), and the second time frame 1b is a window function. w ₂ (t) = 1 (where T−25 ≦ t ≦ T + 25), w ₂ (t) = 0 (otherwise), the fourth time frame 1d is the window function w ₄ (t) = 1 (here T + 25 ≦ t ≦ T + 75) and w ₄ (t) = 0 (otherwise).

Next, amplification performed in the third embodiment will be described. First of all, when the average power L ₁ is higher than the average power L ₂ by a predetermined threshold (5 dB in the second embodiment) or more (that is, L ₁ > L ₂ +5), the amplitude increases in a very narrow range of about 15 ms. Therefore, this increase is regarded as an increase in noise, and L ₁ -L ₂ is calculated to calculate whether it is greater than or less than the threshold. When it is larger than the threshold, the consonant determination unit 4 determines that the voice signal is noise. If it is less than or equal to the threshold value, it is determined according to the following criteria.

Secondly, if L ₁ −L ₂ is equal to or less than the threshold (5 dB) and L ₄ −20 <L ₂ <L ₄ , the consonant determination unit 4 determines that the end point of the consonant or syllable and determines the amplification factor λ Is determined as λ = c · (L ₄ −L ₂ ). Here, c = 0.72. It can be expressed not by the average power difference (L ₄ −L ₂ ) expressed in decibels but also by the ratio of the average powers P ₂ and P ₄ , K ₂₄ = P ₂ / P ₄ . At this time, λ becomes λ = (K ₂₄ ^1/2 ) ^d . d is also a coefficient. The meaning of the coefficient c will be described in Example 4 with reference to FIG. These, when consonants and vowels followed alternating arrangement, since the end point of the consonant or syllable less average power than the vowels, the smaller the L ₂ compares the level of L ₂ level and L _4, the It means that the start point of the consonant or syllable is considered in the 2-hour frame 1b, and the amplification target point or the amplification target width is amplified.

  Note that the amplification target points in Example 3 shown in FIG. 6 are the boundary points between the second time frame 1b and the fourth time frame 1d. As in the first and second embodiments, the audio signal at the center position of the first time frame 1a and the second time frame 1b may be amplified, but when the second time frame 1b and the fourth time frame 1d are provided, Amplifying this boundary can be expected to have an effect. In the third embodiment, this boundary is set as an amplification target point. It is also possible to determine whether to amplify the amplification target point or the amplification target width at the boundary or in the vicinity by arranging the first time frame 1a across both the second time frame 1b and the fourth time frame 1d. . At this time, by providing a fifth time frame (not shown) including the first time frame 1a, it is determined whether or not the audio signal is noise. It is preferable to amplify the target width.

The larger the level difference between the average powers L ₄ and L ₂ displayed in decibels, the larger the amplification, and the difference of 20 dB can be compressed to 5.6 dB. When L ₂ −L ₄ = −20 dB, the amplification degree is maximum at 14.4 dB.

The consonant processing device, the speech information transmission device, and the consonant processing method described above are effective as a consonant enhancement processing device, a speech information transmission device including the consonant enhancement processing device, and a consonant enhancement method, and L ₃ -20 <L ₂ <L. _{Although the} amplification factor λ is increased in the case of ₃ , the amplification factor λ can be made negative by suppressing the amplification. For example, when performing hearing tests, hearing training, etc. for persons with hearing impairment, etc., if the noise continues to be heard for a long time, hearing loss is reduced and uncomfortable feelings. And a consonant suppression processing device and a consonant suppression method that are effective for a listening training apparatus, and voice processing becomes possible.

Furthermore, when L ₁ -L ₂ is equal to or less than the threshold value (5 dB) and L ₂ <L ₄ -20, if L ₂ is 20 dB or more lower than L ₄ , the power is small compared to the previous and next audio signals, and it is impossible. Even if it is amplified, it becomes difficult to distinguish it from noise, so the degree of amplification is gradually reduced. For example, as shown in FIG. 7C, the amplification factor λ is decreased by 3.6 dB every time L ₂ -L ₄ decreases by 10 dB, and the amplification factor λ is set to 0 when L ₂ -L ₃ is −60 dB. is there. However, FIG. 7 (c) is given as an example, and if it is lowered in a stepped manner, the sound becomes uncomfortable where the amplification level changes discontinuously. It is preferable to

  As described above, the third embodiment makes it possible to easily emphasize consonants with a very simple configuration, especially for a CV type language in which important information is often present at the beginning of syllables, such as Japanese and Italian. Become. Note that parameters such as the extraction width, bias, and maximum gain of the time frame 1 can be changed according to the sound environment and the purpose of use.

  Furthermore, the frame divisions of the first and second embodiments are versatile enough to perform consonant enhancement in both Japanese and foreign languages, but the frame division of the third embodiment is effective for consonant enhancement of CV type languages such as Japanese. It can be done. Therefore, the frame division of the third embodiment can be used alone or in combination with the frame division of the first and second embodiments. At this time, the two processes may be performed in parallel to select the one with the larger amplification degree.

  Note that the consonant processing device and the speech information transmission device of the third embodiment can be applied to a VC (Vowel-Consonant) type consonant enhancement process only by reversing the determination of consonants and vowels. When the consonant determination unit 4 determines that it is not the end point of a consonant or syllable, it does not amplify the amplification target point or the amplification target width of the audio signal, but when it is determined as the end point of a consonant or syllable, What is necessary is just to amplify the amplification object width. In this case, the end part of the syllable is emphasized, the end part of the syllable is emphasized, and consonant emphasis of foreign languages other than the CV type language can be effectively performed. Even in Japanese, it is effective for sound-repellent “n”, syllables that are voicing and become silent when uttered.

  Further, when the amplification degree is negative, the voice can be processed into a voice in which the end points of consonants or syllables are difficult to hear, and can be used for hearing test, listening training and the like.

  As described above, the consonant processing device, the speech information transmission device, and the consonant processing method of the third embodiment can perform consonant enhancement only by comparing the difference in average power of frame signals, and thus perform various processes in parallel. It is not necessary, can transmit voice information in a time close to real time, has few logic judgments, is easy to process signals, even under noisy conditions or when the voice competes with other acoustic signals. Even elderly people can easily hear consonants, thereby reducing the strength of the whole voice without impairing the clarity of the voice, and preventing an increase in boundary noise. Moreover, it is suitable for consonant enhancement of CV type languages such as Japanese, and provides a consonant processing device such as a sound enhancement processing device and a voice information transmission device that can be manufactured at low cost because the amplification degree can be easily adjusted. can do. If it is determined that it is not the end point of a consonant or syllable, the amplification target point or amplification target width of the audio signal is not amplified, and if it is determined that it is an end point of a consonant or syllable, the amplification target point or amplification target width of the audio signal , The end of the syllable is emphasized, and consonant enhancement such as foreign languages other than the CV type language can be effectively performed. Also, when used as a consonant suppression processing device, the speech is used for hearing test, listening training, etc. Can be processed for.

Example 4
A consonant processing device, a voice information transmission device, and a consonant processing method according to Embodiment 4 of the present invention will be described. FIG. 8 is an explanatory diagram of the amplification characteristics of the consonant processing device according to the fourth embodiment of the present invention, FIG. 9 is an explanatory diagram of a voice stimulus pattern, and FIG. 10 is a comparison diagram of correct answer rates before and after consonant enhancement processing for each voice stimulus.

The comparison unit 3 of the third embodiment calculates the amplification degree by calculating the difference between the average power L _i (i = 1, 2, 4) displayed in decibels. In the fourth embodiment, the comparison unit 3 calculates the average power P _i ( The ratio is calculated by calculating the ratio of i = 1, 2, 4). Therefore, the same reference numerals are used in the fourth embodiment and the third embodiment, and the other points are the same as those in the third embodiment except that the calculation method of the comparison unit 3 is different. These detailed explanations are given in Example 3 and are omitted here. Therefore, reference is made to FIGS.

In the fourth embodiment, the comparison unit 3 uses the ratio K _ij = P _i / P _j (i, j = 1,2,4) of the average power P _i (i = 1,2,4) of each frame signal; i < j) is calculated, and the amplification degree determination unit 7 calculates the amplification degree. As for the relationship between L ₁ and L ₂ , it is sufficient that noise can be discriminated by a ratio as in the third embodiment. Therefore, the relationship between L ₂ and L ₄ will be described below.

Comparator 3, calculates a second decibels the L ₂ which is the output from the power calculation unit _2b, the difference L ₂ -L ₄ of the fourth decibels the L ₄ is the output from the power calculating unit 2d, L _{If 2-} L ₄ > 0, the amplification degree determination unit 7 does not perform amplification. When the determination is made with the ratio K ₂₄ = P ₂ / P ₄ , K ₂₄ > 1. This is a case where L is larger than point A in FIG.

On the other hand, when L ₂ −L ₄ ≦ 0 or the ratio K ₂₄ ≦ 1, the amplification degree determination unit 7 performs amplification. In this case, the amplification degree is λ = c · (L ₂ −L ₄ ). In FIG. 8, this c is represented by a ratio represented by (line segment βγ) / (line segment αγ). Increasing c increases the degree of amplification, and c
When is 0, the audio signal is not amplified. If there is an input of a γ point on the broken line, if L ₂ <L ₄ , it means that the output is lifted by the line segment βγ and amplified to the β point.

  In FIG. 8, point B is a knee point (amplification degree switching point), and an input / output signal at a level below this becomes difficult to distinguish from noise, so the amplification degree is lowered. In the case of FIG. 8, the knee point B point is set to −20 dB, and for input / output signals smaller than the knee point B point, the amplification degree is gradually decreased, and the amplification degree is maximized at the knee point B point.

Further, in the consonant processing device and the speech information transmission device according to the fourth embodiment, the average power difference L ₂ −L ₄ expressed in decibels is L ₂ −L ₄ > 0, L ₂ −L ₄ ≦ 0, or the average power P _2. , P ₄ ratios K ₂₄ > 1 and K ₂₄ ≦ 1, for example, can be applied to VC-type consonant enhancement processing. That is, by inverting the determination of the amplification target point or the amplification degree of the amplification target width, the terminal part of the syllable is emphasized, and consonant enhancement of foreign languages other than the CV type language can be effectively performed. This frame division can also be used in combination with the frame divisions of the first, second, and third embodiments. When combining, two processes should be performed in parallel and the one with the larger amplification degree may be selected. Thus, the ratio K ₂₄ amplifies the audio signal if 1 or less when it is determined that the final end of the syllable, it is possible to clarify the last syllable.

  Now, in order to confirm the effectiveness of the consonant processing device of Example 4, the intelligibility was verified. As a voice stimulus, a sound source for adult consonant examination recorded in “Evaluation test for speech listening by wearing a cochlear implant (CI2004)” (edited by the Japan Cochlear Implant Research Group) was used. The sound source includes “aba”, “ada”, “aga”, “aha”, “aka”, “ama”, “ana”, “apa”, “ara”, “asa”, “ata”, “ Fourteen types of VCV syllables “awa”, “aya”, and “aza” are provided. This sound source was prepared at 44.1 kHz and subjected to consonant emphasis processing and not processed, and as shown in FIG. 9, background noise having upper and lower limit frequencies of 8000 Hz and 50 Hz was added to obtain a voice stimulus. The background noise has a duration of 5000 ms and a rise and fall of 500 ms, and a voice stimulus subjected to consonant enhancement processing is arranged at the center of the duration of 5000 ms. The time interval until the next voice stimulus was 2000 ms.

  This voice stimulus was given to 14 experimental participants with normal hearing ability, and the correct answer rates were compared between those subjected to consonant enhancement processing and those not processed. FIG. 10 compares the correct answer rate before and after the consonant enhancement processing for each voice stimulus. Comparing the average value of both of them in FIG. 10, it can be seen that the one subjected to the consonant enhancement process is higher than the one not processed. It turns out that the consonant processing apparatus of Example 1 is functioning effectively.

  Among them, the correct answer rate of “asa” and “aza” is particularly high because most of the energy of “s” and “z” is over 8000 Hz and was not masked by background noise. S and “z” are considered to have more audio information in the noise part (consonant part) than in this transition part. Therefore, it is considered that the consonant enhancement processing by the amplifying unit 6 of Example 1 functions effectively, and the clarity is increased.

  On the other hand, “aba”, “ada”, and “aga” are voiced closed consonants, and the shape of the second formant transition has a great influence on the voice identification. The voiced burst consonant is considered to have a lot of voice information in the second formant transition, and the amplitude of the second formant transition part shows a large value with respect to the vowel. It is considered that there was not much difference in clarity between those with and without the consonant enhancement processing.

  As described above, the consonant processing device, the speech information transmission device, and the consonant processing method of the fourth embodiment can perform consonant enhancement only by comparing the ratios of the average powers of the frame signals. Therefore, it is not necessary to perform various processes in parallel. Voice information can be transmitted in a time close to real time, signal processing is simple, and consonants are easy to hear even in deaf and elderly people, even under noisy conditions or when the voice competes with other acoustic signals. Thus, the strength of the entire voice can be reduced without impairing the clarity of the voice, and an increase in environmental noise can be prevented. Also suitable for consonant emphasis in CV type languages such as Japanese, and providing a consonant processing device such as a consonant emphasis processing device and a voice information transmission device that can be manufactured at low cost because the amplification level can be easily adjusted. can do.

  Further, when the amplification degree is negative as in the third embodiment, the speech can be processed into speech in which consonant or syllable end points are difficult to hear as a consonant suppression processing device or the like, which can be used for hearing test, listening training, and the like.

(Example 5)
Hereinafter, a consonant processing device, a voice information transmission device, and a consonant processing method according to Embodiment 5 of the present invention will be described. FIG. 11 is a configuration diagram of a consonant processing device and a voice information transmission device equipped with the same in Embodiment 5 of the present invention.

  The consonant processing apparatus according to the fifth embodiment processes a speech signal in advance and inputs it into the time frame 1 in order to detect the consonant or syllable boundary of the speech signal more clearly.

  Reference numeral 8 shown in FIG. 11 denotes a filter unit placed immediately before the frame dividing unit 1. The filter unit 8 has a characteristic that allows a frequency component of 3000 Hz or less to pass therethrough and has a peak in the vicinity of 1000 Hz, which makes it possible to more appropriately detect a boundary between consonants or syllables. In addition, although Example 5 has shown what provided the filter part 8 in the consonant processing apparatus 10 and the audio | voice information transmission apparatus 20 of Example 1, the filter part 8 and the consonant processing apparatus 10 of Examples 2-4 are shown. The same applies to the voice information transmission device 20. These are not shown.

  As described above, the consonant processing device, the voice information transmitting device, and the consonant processing method of the fifth embodiment can easily detect the boundary between consonants or syllables easily, and the noise or the voice competes with other acoustic signals. Even in situations, it is easy for the hearing impaired and elderly people to hear.

  The present invention can be applied to audio information transmission devices such as hearing aids, such as announcement broadcasting devices and mobile phones.

1 is a configuration diagram of a consonant processing device and a voice information transmission device equipped with the same in Embodiment 1 of the present invention. Explanatory drawing of the process of the consonant processing apparatus in Example 1 of this invention The block diagram of the consonant processing apparatus in Example 2 of this invention and the audio | voice information transmission apparatus carrying this (A) Explanatory diagram of replenishment phenomenon, (b) Comparison diagram of sound in quiet environment and sound in noise The block diagram of the consonant processing apparatus in Example 3 of this invention and the audio | voice information transmission apparatus carrying this Explanatory drawing of the process of the consonant processing apparatus in Example 3 of this invention (A) Explanatory diagram of the amplification factor in the first embodiment of the present invention, (b) Explanatory diagram of the amplification factor in the second embodiment of the present invention, (c) Amplification in the third embodiment of the present invention Illustration of the degree of amplification Explanatory drawing of the amplification characteristic of the consonant processing apparatus in Example 4 of this invention Explanation of voice stimulation pattern Comparison of correct answer rate before and after consonant enhancement for each voice stimulus Configuration diagram of consonant processing device and voice information transmission device equipped with the same in Embodiment 5 of the present invention

Explanation of symbols

DESCRIPTION OF SYMBOLS 1 Frame division | segmentation part 1a 1st time frame 1b 2nd time frame 1c 3rd time frame 1d 4th time frame 2 Power calculation part 2a 1st power calculation part 2b 2nd power calculation part 2c 3rd power calculation part 2d 4th power Calculation unit 3 Comparison unit 4 Consonant determination unit 5 Delay unit 6 Amplification unit 7 Amplification degree determination unit 8 Filter unit 10 Consonant processing device 11 Microphone 12 Speaker 20 Audio information transmission device

Claims (16)

A frame dividing unit that extracts a frame signal from each of a plurality of time frames from an input audio signal;
A power calculator that calculates an average power for each of the frame signals;
A comparison unit for comparing average powers between the frame signals;
A consonant determination unit that determines whether an amplification target point or an amplification target width of the audio signal is an end point of a consonant or a syllable based on a comparison result of the comparison unit;
When the consonant determination unit determines that it is an end point of a consonant or syllable, the amplification unit includes an amplification unit that amplifies the amplification target point or the amplification target width of the audio signal and does not amplify when it is determined that it is not the end point of a consonant or syllable A consonant processing apparatus characterized by that. A frame dividing unit that extracts a frame signal from each of a plurality of time frames from an input audio signal;
A power calculator that calculates an average power for each of the frame signals;
A comparison unit for comparing average powers between the frame signals;
A consonant determination unit that determines whether an amplification target point or an amplification target width of the audio signal is an end point of a consonant or a syllable based on a comparison result of the comparison unit;
When the consonant determination unit determines that it is an end point of a consonant or syllable, it determines the amplification target point or amplification target width of the audio signal in the amplification direction, and when it determines that it is not an end point of a consonant or syllable An amplification degree determination unit that determines that the signal is not amplified;
A consonant processing apparatus comprising: an amplification unit that amplifies the audio signal in accordance with the amplification degree determined by the amplification degree determination unit. 3. The consonant processing device according to claim 1, wherein the comparison unit performs comparison by calculating a difference in decibel average power of each frame signal. 4. The consonant processing apparatus according to claim 1, wherein the comparison unit compares the frame signals by calculating a ratio of average power of the frame signals. The time frame is provided with a time frame having an extraction width capable of extracting a consonant, and the amplification target point or the amplification target width is set at a center position of the extraction width of the time frame. 4. The consonant processing device according to any one of 4 above. The amplification target point or the amplification target width is set at a boundary between the two time frames when two time frames continuous to the time frame are provided. The consonant processing device described in 1. When the difference of the average power displayed in decibels is 0 or less, the amplitude of the audio signal at the amplification target point or the width of the amplification target is amplified, and when the difference displayed in decibels is larger than 0, it is not amplified. The consonant processing device according to claim 3. When the ratio between the average powers is 1 or less, the amplitude of the audio signal at the amplification target point or the amplification object width is amplified, and when the average power ratio is greater than 1, the amplification is not performed. Item 4. The consonant processing device according to Item 4. 7. The consonant processing device according to claim 1, wherein the amplifying unit amplifies the amplification target point or the amplification target width of the audio signal when the consonant determination unit determines that it is an end point of a consonant or a syllable. Instead, the consonant processing device is characterized in that when it is determined as an end point of a consonant or a syllable, the amplification target point or the amplification target width of the audio signal is conversely suppressed. The consonant processing device according to any one of claims 2 to 6, wherein the amplification degree determination unit amplifies an amplification target point or an amplification target width of the audio signal when the consonant determination unit determines that it is an end point of a consonant or a syllable. Instead of determining the degree of amplification in the amplification direction, the amplification degree determination unit reverses the amplification degree of the amplification target point or the amplification target width of the audio signal when the consonant determination unit determines that it is an end point of a consonant or a syllable. A consonant processing device, characterized in that a determination is made that the direction is suppressed. The consonant processing apparatus according to claim 1, further comprising a filter unit that allows a predetermined frequency component to pass before an audio signal is input to the frame dividing unit. The amplifying unit amplifies a physical sound pressure in accordance with a correction characteristic of a supplementary phenomenon that matches a loudness that is a sensory amount of a hearing-impaired hearing person with a loudness that is a sensory amount of a normal hearing person. The consonant processing apparatus according to claim 1, wherein the consonant processing apparatus is characterized. An audio comprising: the consonant processing device according to any one of claims 1 to 12; and a speaker that outputs a consonant-enhanced sound based on a consonant-processed audio signal from the consonant processing device. Information transmission device. A frame signal is extracted for each of a plurality of time frames from the input audio signal, an average power is calculated for each of the frame signals, and the average power is compared between the frame signals. It is determined whether the point to be amplified or the width to be amplified is an end point of a consonant or syllable. A consonant processing method characterized by not amplifying if it is determined not to be a consonant. 15. The consonant processing method according to claim 14, wherein when it is determined as an end point of a consonant or a syllable, instead of amplifying an amplification target point or an amplification target width of the audio signal, it is determined as an end point of a consonant or a syllable Is a consonant processing method characterized by conversely suppressing the amplification target point or the amplification target width of the audio signal. The physical sound pressure is amplified in accordance with a correction characteristic of a supplementary phenomenon that matches the loudness of a sensory sensory hearing person with the loudness of a normal hearing person. The consonant processing method described in 14 or 15.

Images