KR100554165B1 - CELP-based Speech Codec capable of eliminating of pitch-multiple effect and method of the same - Google Patents

Abstract

Disclosed are a CELP-based speech coder capable of eliminating the effects of multiple pitch delay values and a method of eliminating the effects of multiple pitch delay values. The first pitch connection unit calculates a first pitch delay value corresponding to the current pitch delay value from the input voice signal. The storage unit stores the first pitch delay value and 1 / n (n = 2,3) of the first pitch delay value. The comparison unit compares the previous pitch delay value with the 1 / n value of the first pitch delay value. The second pitch connection unit sets the first pitch delay value to the previous pitch delay value if the at least one of the previous pitch delay value and the 1 / n value of the first pitch delay value is the same, and the previous pitch delay value and the first pitch value. If the 1 / n values of the pitch delay values are not all the same, the first pitch delay value is set as the current pitch delay value. According to the present invention, the sound quality can be improved by eliminating the influence of the integer multiple of the pitch delay value during speech encoding by the speech encoder or during mutual encoding of the speech encoding period.

CELP, speech coder, pitch delay value, multiple effect

Description

CELP-based Speech Codec capable of eliminating of pitch-multiple effect and method of the same}

1 is a diagram showing a pitch delay value that is an integer multiple of an actual pitch delay value,

2 is a diagram comparing only the integer value of the pitch delay value determined when the same voice input is encoded through the G.729A speech encoder and the integer value of the pitch delay value determined in the intermediate process when the SMV speech encoder is encoded.

3 is a block diagram showing the configuration of a preferred embodiment of a CELP-based speech coder capable of removing the influence of multiples of a pitch delay value according to the present invention;

4 is a diagram illustrating a result of removing a pitch doubling or pitch triple effect for the same voice; and

FIG. 5 is a flowchart illustrating an embodiment of a method of canceling an influence caused by a multiple of a pitch delay value in a voice encoder according to the present invention.

The present invention relates to a CELP-based speech coder capable of removing the effects of multiple pitch delay values and a method of removing the effects of multiple pitch delay values. More particularly, the present invention relates to a pitch of an integer multiple of an actual pitch delay value. The present invention relates to a CELP-based speech encoder capable of preventing degradation of sound quality due to a delay value and a method for removing the influence by multiples of a pitch delay value.

A speech encoder of a Code Excited Linear Prediction (CELP) structure plots speech by passing arbitrary excitation signals through a linear prediction filter. The linear prediction filter is made by linear prediction analysis, and the excitation signals are generated through analysis by synthesis. The excitation signal is classified into an adaptive codebook and a fixed codebook.

The adaptive codebook consists of adaptive codebook gain and pitch delay value. In the CELP-structured speech encoder, the pitch delay value represents a periodic characteristic of speech. The speech encoder obtains a pitch delay using a normalized correlation. The pitch delay value is mainly expressed in the form of an integer value + a decimal value. The pitch delay value has a different integer range according to the voice encoder, and the range of the decimal number is also different.

Some of the CELP-based speech encoders do not find an actual pitch delay value, but find a pitch delay value that is an integer multiple of the actual pitch delay value. The phenomenon in which the pitch delay value that is an integer multiple of the actual pitch delay value appears is called pitch doubling or pitch triple ring. 1 illustrates a pitch delay value 110 that is an integer multiple of an actual pitch delay value 100 and an actual pitch delay value. An integer multiple of the pitch delay value compared to the actual pitch delay value causes a constant deterioration in sound quality.

The technical problem to be achieved by the present invention is a CELP-based speech coder capable of preventing the degradation of sound quality due to a pitch delay value corresponding to an integer multiple of the actual pitch delay value shown in the speech coder and a method for removing the influence by multiples of the pitch delay value. To provide.

In order to achieve the above technical problem, the CELP-based speech coder capable of removing the influence of multiples of the pitch delay value according to the present invention calculates a first pitch delay value corresponding to a current pitch delay value from an input speech signal. A first pitch connection; A storage unit for storing the first pitch delay value and 1 / n (where n = 2,3) of the first pitch delay value; A comparison unit comparing a previous pitch delay value with a 1 / n value of the first pitch delay value; And setting the first pitch delay value to a previous pitch delay value if the at least one of the previous pitch delay value and the 1 / n value of the first pitch delay value is the same, and setting the first pitch delay value and the first pitch delay value. And a second pitch connection unit configured to set the first pitch delay value to a current pitch delay value if all of 1 / n values of the pitch delay values are not equal to each other.

In order to achieve the above another object of the present invention, the method for removing the influence by the multiple of the pitch delay value in accordance with the present invention, in the method of eliminating the effect of the multiple of the pitch delay value in the CELP-based speech coder or mutual encoder, (a) calculating a first pitch delay value corresponding to a current pitch delay value from the input voice signal; (b) storing the first pitch delay value and the 1 / n value (n = 2,3) of the first pitch delay value in a memory; (c) comparing a previous pitch delay value with a 1 / n value of the first pitch delay value; And (d) if at least one of the previous pitch delay value and the 1 / n value of the first pitch delay value is the same, set the first pitch delay value to a previous pitch delay value, and And setting the first pitch delay value to a current pitch delay value if all of 1 / n values of the first pitch delay value are not equal to each other.

Thereby, the sound quality can be improved by eliminating the influence of the integer multiple of the pitch delay value during speech encoding by the speech encoder or during mutual encoding of the speech encoding period.

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS Hereinafter, a preferred embodiment of a CELP-based speech coder capable of canceling the influence of the pitch delay value by a multiple of the pitch delay value and the method of removing the influence of the pitch delay value according to the present invention will be described in detail.

FIG. 2 is a diagram comparing only integer values of pitch delay values determined when encoding the same voice input through a G.729A voice encoder and integer values of pitch delay values determined in an intermediate process when encoding through an SMV voice encoder. . In the area 200 indicated by the circle, the pitch delay value for G.729A appears to be quite large. This is because pitch doubling of the G.729A voice encoder is not properly performed, resulting in pitch doubling or pitch triple ringing.

Pitch doubling or pitch triplering occurs mainly when the pitch delay value is kept constant, i.e. when the voiced sound section continues constant. The process of obtaining the pitch delay value is a process of determining the periodic characteristics of the voice. The pitch doubling or the pitch triple occurs when the voice encoder obtains a period different from the period of the voice actually input when the signal has a constant period. That is, since T may be a period of 2T if the period is a characteristic of the period, the phenomenon in which the voice encoder determines 2T as the period of the input voice in the process of obtaining the period may be referred to as pitch doubling or pitch tripleling. .

For this reason, the pitch doubling or pitch triple phenomena occur, so if the pitch delay value is constant and suddenly doubles or triples the existing pitch delay value, it may be determined that the pitch delay value is incorrectly determined. Theoretically, four or five times the values may come out, but these values are not really good.

3 is a block diagram showing the configuration of a preferred embodiment of a CELP-based speech coder capable of removing the influence of multiples of the pitch delay value according to the present invention.

Referring to FIG. 3, the CELP-based speech coder capable of removing the influence of a multiple of the pitch delay value according to the present invention includes a framer 310, a high pass filter 320, a linear predictive coefficient extractor 330, and a linear predictor. A filter 340, a low pass filter 350, and a pitch delay calculation unit 360 are provided.

The framer 310 performs framing to analyze the input voice signal. The framer 310 divides the voice signal continuously received through the PCM line into an analysis section defined in the voice encoder. The high pass filter 320 removes unwanted low frequency components included in the audio signal from the output signal of the framer 311. The linear prediction coefficient extractor 330 extracts a linear prediction coefficient (LPC) from the high pass filtered signal. The linear prediction filter 340 removes the formant component of the voice signal by using the LPC extracted by the linear prediction coefficient extractor 330. The low pass filter 350 filters components other than the pitch in the linear predictive filtered signal. The pitch delay calculation unit 360 calculates a pitch delay value from the low pass filtered signal, and determines a final pitch delay value by removing an integer multiple of the pitch delay value based on the calculated pitch delay value.

The pitch delay calculation unit 360 includes a first pitch connection unit 362, a storage unit 364, a comparison unit 366, and a second pitch connection unit 368. The first pitch connection unit 362 calculates a first pitch delay value corresponding to the current pitch delay value from the low pass filtered signal. The storage unit 364 stores the calculated first pitch delay value, ½ value of the first pitch delay value, and the? Value of the first pitch delay value. The comparator 366 compares the 1/2 value of the previous pitch delay value and the first pitch delay value, and compares the? Value of the previous pitch delay value and the first pitch delay value. The second pitch connection unit 368 resets the pitch delay value based on the comparison result of the comparison unit 366. The second pitch connection unit 368 sets the first pitch delay value to the previous pitch delay value when the ½ or ⅓ value of the previous pitch delay value and the first pitch delay value are the same. On the contrary, if the pitch value of the first pitch delay value and the first pitch delay value are not equal to each other, the second pitch connection unit 368 replaces the first pitch delay value with the current pitch delay value. Set to.

On the other hand, when converting the pitch delay value in the inter encoder, instead of performing open loop pitch detection, a closed loop pitch detection process based on the pitch delay value of the input of the inter encoder. Do this. By using the present invention, it is possible to prevent the problem that the pitch delay value is incorrectly determined in the closed loop search process. In the case where the present invention is applied to the mutual encoder, the closed loop pitch search unit 370 is provided, and the closed loop pitch search unit 370 is indicated by a dotted line in FIG. 3. The closed loop pitch search unit 370 performs a closed loop pitch search based on the first pitch delay value determined by the first pitch connection unit 362, changes the first pitch delay value, changes the first pitch delay value, The ½ value of the changed first pitch delay value and the ⅓ value of the changed first pitch delay value are stored in the storage unit 364. In this case, the framer 310, the high pass filter 320, the linear predictive coefficient extractor 330, the linear predictive filter 340, the low pass filter 350, and the first pitch connection unit 362 are sheared. Configure a voice decoder. The closed-loop pitch search unit 370, the storage unit 364, the comparison unit 366, and the second pitch connection unit 368 are provided in a speech encoder at a later stage to determine a pitch delay value when mutually encoding the speech. do.

4 is a diagram illustrating a result of removing a pitch doubling or pitch triple effect for the same voice, and it can be seen that values corresponding to integer multiples of the pitch delay value are removed in the region 400 indicated by the circle.

FIG. 5 is a flowchart illustrating an embodiment of a method of canceling an influence caused by a multiple of a pitch delay value in a voice encoder according to the present invention.

Referring to FIG. 5, the first pitch connection unit 362 calculates a first pitch delay value corresponding to a current pitch delay value from the low pass filtered signal (S500). The calculated first pitch delay value, the half value of the first pitch delay value, and the value of the first pitch delay value are stored in the storage unit 364 (S510). On the other hand, in the speech inter coder, step S500 is performed by the voice decoder of the front stage, and the voice encoder of the rear stage performs the closed loop pitch search based on the calculated first pitch delay value to calculate the first pitch delay value calculated in step S500. In operation S520, the changed first pitch delay value is stored in the storage unit 364. The step S520, which is performed only in the voice encoder, is shown by a dotted line. The comparator 366 compares the ½ value of the previous pitch delay value and the first pitch delay value (S530), and compares the value of the previous pitch delay value and the first pitch delay value (S540). The second pitch connection unit 368 sets the first pitch delay value to the previous pitch delay value if the previous pitch delay value and the ½ or ⅓ value of the first pitch delay value are the same (S550). On the contrary, if the second pitch connection unit 368 does not have the same value as ½ and ⅓ of the previous pitch delay value and the first pitch delay value, the second pitch connection unit 368 sets the first pitch delay value as the current pitch delay value (S560). .

The CELP-based speech coder capable of removing the influence of multiple pitch delay values and the method of removing the influence of multiple pitch delay values according to the present invention may be applied to a general voice encoder, not a mutual encoder. In addition, although the present invention can be applied to both a decoder and an encoder in a speech encoder, the present invention is generally applied to an encoder. Therefore, after the encoder determines the range of the pitch delay value through the open circuit search, the pitch delay value is corrected by removing the influence of the multiple of the pitch delay value as described above.

The invention can also be embodied as computer readable code on a computer readable recording medium. Computer-readable recording media include all kinds of recording devices that store data that can be read by a computer device. Examples of computer-readable recording media include ROM, RAM, CD-ROM, magnetic tape, floppy disk, optical data storage, and the like, and may also be implemented in the form of a carrier wave (for example, transmission over the Internet). Include. The computer readable recording medium can also be distributed over network coupled computer devices so that the computer readable code is stored and executed in a distributed fashion.

Although the preferred embodiments of the present invention have been shown and described above, the present invention is not limited to the specific preferred embodiments described above, and the present invention belongs to the present invention without departing from the gist of the present invention as claimed in the claims. Various modifications can be made by those skilled in the art, and such changes are within the scope of the claims.

According to the present invention, a CELP-based speech coder capable of removing the influence of multiple pitch delay values and a method for removing the influence of multiple pitch delay values may include a pitch delay during speech encoding by a speech encoder or mutual encoding of a speech encoding period. The sound quality can be improved by removing the influence of the integer multiple of the value.

Claims (4)

A first pitch connection unit for calculating a first pitch delay value corresponding to a current pitch delay value from the input voice signal; A storage unit for storing the first pitch delay value and 1 / n (where n = 2,3) of the first pitch delay value; A comparison unit comparing a previous pitch delay value with a 1 / n value of the first pitch delay value; And If at least one of the previous pitch delay value and the 1 / n value of the first pitch delay value is the same, the first pitch delay value is set as a previous pitch delay value, and the previous pitch delay value and the first pitch If the 1 / n value of the delay value is not all the same, the second pitch connection unit for setting the first pitch delay value to the current pitch delay value; comprising: CELP-based speech coder available. The method of claim 1, The first pitch delay value is changed by performing a closed loop pitch search based on the first pitch delay value determined by the first pitch connection unit, and the changed first pitch delay value and the changed first pitch delay value are determined. A closed loop pitch search unit for storing the 1 / n value of the storage unit; CELP-based speech coder that can remove the effect of multiple times the pitch delay value further comprises. In the method of eliminating the effect of multiples of the pitch delay value in the CELP-based speech coder or mutual encoder, (a) calculating a first pitch delay value corresponding to a current pitch delay value from the input voice signal; (b) storing the first pitch delay value and the 1 / n value (n = 2,3) of the first pitch delay value in a memory; (c) comparing a previous pitch delay value with a 1 / n value of the first pitch delay value; And (d) if at least one of the previous pitch delay value and the 1 / n value of the first pitch delay value is the same, set the first pitch delay value to a previous pitch delay value, and the previous pitch delay value and the And setting the first pitch delay value as a current pitch delay value if all of 1 / n values of the first pitch delay value are not equal to each other. The method of claim 3, wherein In step (a), (a1) calculating a first pitch delay value corresponding to a current pitch delay value from the input voice signal; (a2) changing the first pitch delay value by performing a closed loop pitch search based on the calculated first pitch delay value; And (a3) storing the changed first pitch delay value and the 1 / n value of the changed first pitch delay value in the memory.

Images