JP6451079B2 - Speech enhancement device and program, and speech decoding device and program - Google Patents

Description

  The present invention relates to a speech enhancement device and program, and a speech decoding device and program, and can be applied to, for example, a call system.

  In recent years, digitalization of call systems has progressed, and most call systems such as mobile phone systems have been digitized. In such a digital call system, a voice compression technique (voice coding technique) is used mainly for the purpose of reducing the amount of data on the transmission path. In general, the quality of an audio signal obtained by encoding and transmitting speech on the transmitting side and decoding the encoded audio signal on the receiving side is deteriorated compared to the audio signal before being encoded. In particular, fine changes in the sound level are averaged, resulting in a decrease in clarity. For such a problem, for example, it is conceivable to apply a speech enhancement process as described in Patent Document 1 to a speech signal. The speech enhancement technique described in Patent Literature 1 enhances the clarity by enhancing the level change at the rising portion of the speech signal.

Japanese Patent Laid-Open No. 07-104788

  By the way, the audio signal obtained by the decoding process on the encoded audio signal is a digital signal. On the other hand, the speech enhancement device described in Patent Document 1 uses a plurality of time constant circuits. The time constant circuit is a circuit suitable for realization with an analog circuit.

  Since the decoded audio signal is a digital signal, it may be considered that the time constant circuit is constituted by a digital circuit. However, in addition to realizing a time constant circuit with a complex circuit such as a digital filter, a speech enhancement device that enhances a speech signal of a digital signal becomes complicated and large. It is also conceivable that the decoded speech signal of the digital signal is converted into an analog signal and input to the speech enhancement device described in Patent Document 1. However, in addition to speech enhancement, the decoded speech signal is often subjected to other processing such as an equalizer, and it is wasteful in configuration to configure only a part of the circuit with an analog circuit.

  Nowadays, it is also conceivable to perform processing performed by digital processing by software processing. However, when performing software processing, the program corresponding to the time constant circuit becomes complicated, or even if other processing is realized by software processing, the configuration must be configured by an analog circuit for speech enhancement. Etc. have the same problems as described above.

  Therefore, a speech enhancement device and program suitable for digital processing and software processing, and a speech decoding device and program to which such speech enhancement device and program are applied are desired.

According to a first aspect of the present invention, there is provided a speech enhancement apparatus in which a gain enhancement unit multiplies a speech signal by gain to enhance (1) a sample level calculation unit that calculates a level for each sample of the speech signal, and (2) a predetermined sample. Representative value calculating means for calculating a representative value of the sample level in a predetermined number of samples based on a number of sample levels; and (3) the voice based only on the representative value of the processing target section and the representative value of the immediately preceding section. And gain calculating means for calculating a gain for enhancing the signal.

The second aspect of the present invention is a speech enhancement program for emphasizing a speech signal, comprising: (1) sample level calculation means for calculating a level for each sample of the speech signal; and (2) a sample level of a predetermined number of samples. And (3) enhancing the audio signal based only on the representative value of the section to be processed and the representative value of the immediately preceding section. And (4) a gain enhancement unit that multiplies the audio signal by the gain for enhancement.

  A third aspect of the present invention is a speech decoding apparatus having a speech decoding unit that decodes an encoded speech signal and a speech enhancement unit that emphasizes the decoded speech signal, and the first book is used as the speech enhancement unit. The speech enhancement device of the invention is applied.

  The fourth aspect of the present invention is a speech decoding program that causes a computer to function as a speech decoding unit that decodes an encoded speech signal and a speech enhancement unit that enhances the decoded speech signal, and functions as the speech enhancement unit. The speech enhancement program according to the second aspect of the present invention is applied as the program portion.

  According to the present invention, it is possible to realize a speech enhancement device and program suitable for digital processing and software processing, and a speech decoding device and program to which such speech enhancement device and program are applied.

It is a block diagram which shows the functional structure of the speech decoding apparatus which concerns on 1st Embodiment. It is a functional block diagram which shows the internal structure of the audio | voice emphasis part (voice | voice emphasis apparatus) in the audio | voice decoding apparatus of 1st Embodiment. It is a functional block diagram which shows the internal structure of the speech enhancement part (speech enhancement apparatus) in the speech decoding apparatus of 2nd Embodiment. It is explanatory drawing of the 1st restriction | limiting method of the gain for emphasis in the audio | voice emphasis part of 2nd Embodiment. It is explanatory drawing of the 2nd restriction | limiting method of the gain for emphasis in the audio | voice emphasis part of 2nd Embodiment. It is explanatory drawing of the 3rd restriction | limiting method of the gain for emphasis in the audio | voice emphasis part of 2nd Embodiment.

(A) First Embodiment Hereinafter, a first embodiment of a speech enhancement device and program, and a speech decoding device and program according to the present invention will be described with reference to the drawings.

(A-1) Configuration of First Embodiment FIG. 1 is a block diagram showing a functional configuration of a speech decoding apparatus according to the first embodiment. Here, the speech decoding apparatus according to the first embodiment can be configured by hardware, and can also be realized by software (speech decoding program) executed by the CPU and the CPU. Whichever implementation method is employed, it can be functionally represented in FIG.

  In FIG. 1, the speech decoding apparatus 1 according to the first embodiment includes a speech decoding unit 2 and a speech enhancement unit 3.

  The speech decoding unit 2 receives an encoded speech signal that is obtained by demodulating a transmission path by a receiving unit (not shown). The audio decoding unit 2 decodes the encoded audio signal, obtains a decoded audio signal (digital signal), and gives it to the audio enhancement unit 3. Here, the speech coding method is not limited, but the effect of the first embodiment becomes remarkable in the case of a high-efficiency coding method with a high compression rate such as a CELP (Code Excited Linear Prediction) method. Since the internal configuration of the speech decoding unit 2 is the same as that of the existing one, description of the detailed configuration is omitted.

  The speech enhancement unit 3 corresponds to the speech enhancement device of the first embodiment. The voice emphasizing unit 3 of the first embodiment increases the moment of increase when the decoded voice signal level increases (makes the increase faster), and the momentum of the decrease when the decoded voice signal level decreases. The decoded speech signal is converted into a speech signal that is emphasized so as to increase (decrease the speed). In other words, the voice emphasizing unit 3 emphasizes the level change of the rising part and the falling part of the decoded voice signal.

  FIG. 2 is a functional block diagram showing the internal configuration of the speech enhancement unit 3 of the first embodiment.

  In FIG. 2, the speech enhancement unit 3 includes a level calculation circuit 10, a frame sum calculation circuit 11, a gain calculation circuit 12, a delay circuit 13, and an enhancement operation circuit 14.

  The level calculation circuit 10 calculates a level for each sample of an input decoded audio signal (hereinafter referred to as an input audio signal) S and gives the sample level Sa to the frame sum calculation circuit 11. The absolute value of the input speech signal sample may be calculated as the sample level of the input speech signal, or the square value of the input speech signal sample may be calculated.

  The frame sum calculation circuit 11 calculates the sum of the sample levels Sa for each frame composed of N samples (N is 1 or more) in the input audio signal S, and gives the frame sum Sf to the gain calculation circuit 12. The number N of samples constituting the frame may be any number that can capture the change tendency (increase tendency or decrease tendency) of the sample level Sa based on the frame sum Sf. Although there is a relationship with the sampling rate, the number of samples N may be 1, for example. In addition, the frames that follow each other may be completely separated, and the frame sections that overlap the second half 1 / m sample of the previous frame with the first half 1 / m sample of the rear frame overlap. It may be. The number of samples N and the overlap amount affect the rate of change of increase and decrease described above.

  Although the frame sum calculation circuit 11 of the first embodiment calculates the frame sum as a representative value of the sample level of the frame, another value may be calculated as the representative value. For example, the average value of N sample levels Sa may be used as the frame representative value, and the square root of the frame sum may be used as the frame representative value.

  The gain calculation circuit 12 obtains a value obtained by dividing the frame sum Sf for the frame to be processed by the frame sum (represented by the code Sfp) of the immediately preceding frame, and gains the value g (= Sf / Sfp) as a gain. Is given to the emphasis calculation circuit 14 as follows. If at least one of the frame total Sf of the processing target frame or the frame total Sfp of the immediately preceding frame is 0 and division is not possible, the gain g is set to 1, for example. When division is not possible, the gain g calculated immediately before may be used as the gain for the current processing target frame. Further, instead of the divided value itself, a square root of the divided value or a value obtained by multiplying the divided value by a predetermined value may be used as the gain g. Whether to apply the divided value or the processed value of the divided value affects the change rate of the increase or decrease described above.

  The delay circuit 13 delays the input audio signal S by the processing delay time in the processing system including the level calculation circuit 10, the frame total calculation circuit 11, and the gain calculation circuit 12, and emphasizes the delayed input audio signal Sn. 14. That is, the delay circuit 13 receives the input audio signal S so that the input operation signal 14 is supplied to the enhancement operation circuit 14 at the timing when the gain g of the process target frame is provided to the enhancement operation circuit 14. Is to delay.

  The enhancement operation circuit 14 enhances the input audio signal Sn based on the gain g. For example, the enhancement operation circuit 14 simply multiplies the input audio signal Sn by a gain g to enhance the input audio signal Sn.

Further, the enhancement calculation circuit 14 may change the gain g multiplied by the input audio signal (sample thereof) Sn for each sample so that the change in gain becomes smooth. Formula (1) is a calculation formula for changing the gain for each sample of the input audio signal. In equation (1), g is a gain obtained by calculation for the processing target frame, gp is a gain obtained by calculation for the frame immediately before the processing target frame, and i is the head of all N samples of the processing target frame. Is a parameter that defines the i (i is 1 to N) th input speech signal sample counted from the side, and g (i) is a gain multiplied by the ith input speech signal sample counted from the head side of the processing target frame. Represents.
g (i) = gp * (N−i) / N + g * i / N (1)

(A-2) Operation of the First Embodiment Next, the operation of the speech decoding apparatus 1 of the first embodiment will be described in the order of the overall operation and the speech enhancement operation in the speech enhancement unit 3.

  An encoded speech signal obtained by a receiving unit (not shown) is input to the speech decoding unit 2 and decoded by the speech decoding unit 2. In the obtained decoded speech signal S, the speech enhancement unit 3 emphasizes the level change of the rising portion and the falling portion, and the enhanced speech signal Sg is output.

  Inside the voice emphasizing unit 3, the level change of the rising part and the falling part of the input voice signal (decoded voice signal) S is emphasized as follows.

  The input speech signal S is given to the level calculation circuit 10 and the delay circuit 13 in the speech enhancement unit 3.

  The level for each sample of the input audio signal S is calculated by the level calculation circuit 10, and the obtained sample level Sa is given to the frame sum calculation circuit 11. In the frame sum calculation circuit 11, the sum of all N sample levels Sa is calculated for each frame, and the obtained frame sum Sf is given to the gain calculation circuit 12. In the gain calculation circuit 12, the frame sum Sf for the processing target frame is divided by the frame sum Sfp of the immediately preceding frame, and the quotient g (= Sf / Sfp) is given to the enhancement calculation circuit 14 as a gain.

  On the other hand, the input audio signal S input to the delay circuit 13 is delayed by the processing delay time in the processing system including the level calculation circuit 10, the frame sum calculation circuit 11, and the gain calculation circuit 12, and the delayed input audio signal Sn. Is given to the emphasis calculation circuit 14.

  Then, the emphasis calculation circuit 14 executes an emphasis calculation on the input audio signal Sn based on the gain g, and the audio signal Sg in which the level change of the rising portion and the falling portion of the input audio signal S is emphasized is the enhancement calculation circuit. 14 is output.

  Here, it is assumed that the enhancement operation in the enhancement operation circuit 14 simply multiplies the input audio signal Sn by the gain g. In this case, if the frame sum Sf of the processing target frame is 1.2 times the frame sum Sfp of the immediately preceding frame, the gain g is 1.2. As a result, the value of each sample of the processing target frame of the input audio signal Sn is multiplied by 1.2. On average, in a situation where the value of each sample of the processing target frame is 1.2 times the value of each sample of the immediately preceding frame, the gain g (= 1.2) is further multiplied. The value of each sample after the enhancement calculation is 1.44 (= 1.2 × 1.2) times the value of each sample in the immediately preceding frame. As is clear from the above example, the level change at the rising portion of the input audio signal S is emphasized (from 1.2 to 1.44).

  If the frame sum Sf of the processing target frame is 0.9 times the frame sum Sfp of the immediately preceding frame, the gain g is 0.9. As a result, the value of each sample of the processing target frame of the input audio signal Sn is multiplied by 0.9. On average, in a situation where the value of each sample of the processing target frame is 0.9 times the value of each sample of the immediately preceding frame, the gain g (= 0.9) is further multiplied. The value of each sample after the enhancement calculation is 0.81 (= 0.9 × 0.9) times the value of each sample in the immediately preceding frame. As is clear from the above example, the level change at the falling portion of the input audio signal S is emphasized (from 0.9 to 0.81).

(A-3) Effects of the First Embodiment According to the first embodiment, by simple operations such as absolute value operation (or square operation (multiplication)), summation operation, division, multiplication (or product sum), etc. Thus, it is possible to emphasize the level change of the rising portion and the falling portion of the decoded audio signal (input audio signal).

  By such enhancement processing, even when the level change of the decoded speech signal becomes gentle due to the encoding processing or decoding processing, the level change of the enhanced speech signal becomes clear and the clarity can be improved. it can.

  The speech enhancement described in Patent Document 1 described above can emphasize only the rising portion of the speech signal, but according to the first embodiment, not only the rising portion of the speech signal but also the falling portion of the speech signal. It can be emphasized and the clarity is further improved.

(B) Second Embodiment Next, a second embodiment of the speech enhancement apparatus and program, and the speech decoding apparatus and program according to the present invention will be described with reference to the drawings.

  The speech decoding apparatus 1A according to the second embodiment also includes the speech decoding unit 2 and the speech enhancement unit 3A as shown in FIG. 1 described above, but the speech enhancement unit 3A is different from that of the first embodiment. .

  FIG. 3 is a functional block diagram showing the internal configuration of the speech enhancement unit 3A of the second embodiment, and the same and corresponding parts as those in FIG.

  In FIG. 3, the speech enhancement unit 3 </ b> A according to the second embodiment includes a gain limiting circuit 20 in addition to the level calculation circuit 10, the frame sum calculation circuit 11, the gain calculation circuit 12, the delay circuit 13, and the enhancement operation circuit 14. Since the level calculation circuit 10, the frame total calculation circuit 11, the gain calculation circuit 12, the delay circuit 13, and the emphasis calculation circuit 14 are the same as those in the first embodiment, description of their functions is omitted.

  In the case of the second embodiment, the gain g output from the gain calculation circuit 12 is given to the gain limiting circuit 20.

  When the input gain g is a value within a predetermined range, the gain limiting circuit 20 converts (limits) the output to a value according to the range and outputs the value. The input gain g is a value outside the predetermined range. In this case, the data is output as it is. The gain limiting circuit 20 may be configured with a conversion table, for example. The gain limiting circuit 20 is configured by a digital circuit or software that compares the input gain g with a boundary value (threshold value) of the range and outputs a preset value when belonging to a predetermined range. It may be. The gain limiting circuit 20 gives the gain gl after the limiting process to the emphasis calculation circuit 14. The enhancement calculation circuit 14 performs the same processing as that of the first embodiment based on the gain gl from the gain limiting circuit 20.

  4 to 6 are explanatory diagrams showing examples of relationship between input and output gains of the gain limiting circuit 20 (gain limiting examples). 4 to 6, the horizontal axis represents the value of the gain g input to the gain limiting circuit 20, and the vertical axis represents the value of the gain gl output from the gain limiting circuit 20. Different embodiments constitute different input / output gain relationships, which will be described collectively below.

  In FIG. 4, 1 is output as a limited gain gl for values in the range M to 1 / K (M> 1, K> 1) where the input gain g is close to 1. In other words, a dead zone is introduced into the enhancement process. When the frame sum of the processing target frame and the frame sum of the immediately preceding frame are almost the same and there is almost no level change, that is, when the input gain g is a value close to 1, the enhancement processing is not performed. A gain limit as shown in FIG.

  FIG. 5 introduces a gain upper limit P (P> 1) and a gain lower limit 1 / Q (Q> 1) in addition to the dead zone described with reference to FIG. When the gain g is equal to or higher than the upper limit value P, the upper limit value P is output as the limited gain gl. When the gain g is less than or equal to the lower limit value 1 / Q, the lower limit value 1 / Q is output as the limited gain gl. If the gain is too large or too small, voice enhancement is excessively performed and the sound quality may be deteriorated by the voice enhancement. Therefore, an upper limit value P and a lower limit value 1 / Q are introduced into the gain. FIG. 5 shows a case where both the gain dead zone and the upper and lower limit values of the gain are introduced, but only the limitation of the upper and lower limit values of the gain may be provided.

  FIG. 6 shows the case where the enhancement processing at the falling edge of the audio signal is not executed in addition to the dead zone processing described with reference to FIG. The clarification by speech enhancement is more effective at the rising portion, and such a restriction may be introduced when the processing load is reduced.

  According to the speech enhancement unit (speech enhancement device) of the second embodiment, in addition to the same effects as those of the first embodiment, the gain for speech enhancement is limited. Characteristics, functions, etc. can be realized.

(C) Other Embodiments In the description of each of the above embodiments, various modified embodiments have been mentioned, and further modified embodiments as exemplified below can be given.

  In each of the above embodiments, voice enhancement (including voice enhancement with a gain of 1) is performed in all sections of a voice signal. However, a voice / silence determination is performed, and voice is spoken only for a voice signal in a voice section. Emphasis may be given. Here, depending on the sound / silence determination method, there is a possibility that a part of the rising portion of the audio signal is determined to be a silent section, but an extension period of a predetermined number of samples before and after the determined sound section. May be added to make a voiced section, and it may be ensured that the voice enhancement is appropriately performed at the rising and falling portions of the audio signal. In addition, voice enhancement may be performed only in the first half of a voiced section instead of the whole voiced section. Here, the first half section may have a fixed length, or may have a variable length such as a half section of the sound section.

  In each of the embodiments described above, the processing target section related to the numerator-side feature quantity (for example, frame sum) for calculating the gain and the immediately preceding section related to the denominator-side feature quantity have the same length (frame). The lengths of these sections may be different. For example, the processing target section may be one frame period, and the immediately preceding section may be two frame periods. In this case, if the frame sum is applied as the feature amount, the gain may be calculated using the average of the frame sums of the two frames in the immediately preceding section as the denominator. Further, an average value of sample levels that does not affect the difference in section length between the two sections may be applied as the feature amount.

  In each of the above embodiments, the voice enhancement units 3 and 3A are always operated. However, the user may be able to select whether or not to operate the voice enhancement units 3 and 3A.

  In addition, for example, in the case of a speech decoding apparatus that can support a plurality of speech coding schemes, a speech enhancement unit (or a partial configuration thereof (gain limiting circuit)) corresponding to the speech coding scheme is provided and selected at that time. A voice emphasis unit corresponding to the voice coding method being used may be operated.

  In the second embodiment, the gain limiting circuit always operates. However, the user may select whether to operate the gain limiting circuit.

  Also, as the gain limiting circuit, a plurality of gain limiting circuits having different limiting characteristics (see FIGS. 4 to 6) may be provided so that the user can select the gain limiting circuit to be applied.

  DESCRIPTION OF SYMBOLS 1, 1A ... Speech decoding device, 2 ... Speech decoding unit, 3, 3A ... Speech enhancement unit (speech enhancement device), 10 ... Level calculation circuit, 11 ... Frame total calculation circuit, 12 ... Gain calculation circuit, 13 ... Delay circuit , 14 ... Emphasis calculation circuit, 20 ... Gain limiting circuit.

Claims (9)

In the speech enhancement device in which the gain enhancement unit multiplies the speech signal by gain to enhance the speech signal,
Sample level calculation means for calculating the level of each sample of the audio signal;
Representative value calculating means for calculating a representative value of the sample level in the predetermined number of samples based on the sample level of the predetermined number of samples;
A speech enhancement apparatus comprising: gain calculation means for calculating a gain for enhancing the speech signal based only on the representative value of the processing target section and the representative value of the immediately preceding section.   The speech enhancement apparatus according to claim 1, wherein the sample level calculation means calculates an absolute value or a square value of the sample value as a level for each sample.   The speech enhancement apparatus according to claim 1 or 2, wherein the representative value calculation means calculates a sum of sample levels of a predetermined number of samples as a representative value.   The gain calculation means calculates a quotient obtained by dividing the representative value of the section to be processed by the representative value of the immediately preceding section, or a square root of the quotient as a gain. Voice enhancement device.   5. The apparatus according to claim 1, further comprising gain limiting means for converting the gain calculated by the gain calculating means into a value corresponding to the predetermined range when the gain is a value within the predetermined range. The voice emphasis device described in 1.   The gain emphasizing unit gradually increases, for each sample, a gain for multiplying the sample value of the input audio signal in the processing target section based on the gain in the processing target section and the gain in the section that was the processing target section immediately before. The speech enhancement device according to claim 1, wherein the speech enhancement device is changed to A speech enhancement program for emphasizing speech signals,
Computer
Sample level calculation means for calculating the level of each sample of the audio signal;
Representative value calculating means for calculating a representative value of the sample level in the predetermined number of samples based on the sample level of the predetermined number of samples;
Gain calculation means for calculating a gain for enhancing the audio signal based only on the representative value of the processing target section and the representative value of the immediately preceding section;
A speech enhancement program that functions as a gain enhancement means for enhancing a speech signal by multiplying the gain. A speech decoding apparatus comprising: a speech decoding unit that decodes an encoded speech signal; and a speech enhancement unit that enhances the decoded speech signal,
A speech decoding device, wherein the speech enhancement device according to claim 1 is applied as the speech enhancement unit. An audio decoding program that causes a computer to function as an audio decoding unit that decodes an encoded audio signal and an audio enhancement unit that emphasizes the decoded audio signal,
A speech decoding program, wherein the speech enhancement program according to claim 7 is applied as a program part that functions as the speech enhancement unit.

Images