JP2006039559A - Device and method of audio coding using plp of transfer communication terminal - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide an audio coding method of a transfer communication terminal that can realize higher compressibility than when the traditional LP coefficient is used and ensure higher sound quality by conducting LPC (Linear Predictive Coding) with using PLP (Perceptual Linear Prediction) coefficients. <P>SOLUTION: An audio coding device of the transfer communication terminal includes a PLP coefficient calculating portion 20 processing an input signal and calculating PLP coefficient and gain, a V/UV determination portion 21 determining whether the input signal is a voiced signal or an unvoiced signal and outputting a determination signal and the voiced signal when the input signal is the voiced signal, a pitch calculating portion 22 calculating the pitch of the input signal output from the V/UV determination portion 21 and a parameter coding portion 23 conducting low bit rate coding by using PLP coefficient, gain and pitch based on the determination signal. <P>COPYRIGHT: (C)2006,JPO&NCIPI

Description

  The present invention relates to coding of a mobile communication terminal, and more particularly, to a speech coding apparatus and method using PLP (Perceptual Linear Prediction).

  With the development of mobile communication technology, current mobile communication terminals provide not only voice communication but also data communication using numbers, characters and symbols, and multimedia communication including various video signals. Many terminal users transmit and receive necessary data using radio resources after a radio channel is allocated from the system. However, since the wireless channel has a limited bandwidth so that a large number of users can use it simultaneously, the bit rate of each user is limited.

  Accordingly, a coding technique has been proposed as a technique for transmitting more data at the limited bit rate. There are a wide variety of conventional voice coding techniques, but each voice coding method has predetermined advantages at a predetermined bit rate.

  For example, speech coding using standard audio coding (Generic Audio Coding), PCM (Pulse Code Modulation) and ADPCM (Adaptive Delta Pulse Code Modulation) are effective at a high bit rate of 16 Kbps or more, and P is effective. (Code Excited Linear Prediction) and various modifications thereof are effective at medium bit rates in the range of 2.4 Kbps to 16 Kbps. In particular, at the above-mentioned medium bit rates, LD-CELP (Low Delay Code Excited Linear Prediction), CS-ACELP (Conjugate Structure Excite Linear Excited Linear Prediction), and VSELP (Vex. The coding method used and wideband speech coding are used. In addition, LPC (Linear Predictive Coding), RELP (Residual Excited Linear Prediction), Formant vocoder, Cepstral vocoder (Cepstral Vocoder), etc., which has a bit rate from 75 bps to 2. Have.

  Here, the present invention proposes an improvement method of linear predictive coding (hereinafter referred to as LPC) among codings used at a low bit rate.

  FIG. 3 is a configuration diagram of a conventional LPC encoder.

As shown in FIG. 3, a conventional LPC encoder processes a correlator 10 that calculates an autocorrelation value r _x [n] of an input signal x [n] and the autocorrelation value r _x [n]. Then, LP coefficient calculation unit 11 for calculating linear prediction coefficient (Linear Prediction Coefficient: hereinafter referred to as LP coefficient) a _L and gain G, and whether input signal x [n] is a voice (Voided: V) signal or not A V / UV determination unit 12 that determines whether the signal is a voice (Unvoiced: UV) signal, and a pitch calculation unit 13 that calculates a pitch P of the corresponding signal when the input signal x [n] is a voice (V) signal. And the LP coefficient calculation unit 11 and the pitch calculation unit 1 according to the V / UV instruction bit (Indication bit) output from the V / UV determination unit 12. 3 is configured by a parameter coding unit 14 that codes the LP coefficient a _n , the gain G, and the pitch P received from 3 and outputs a bit stream.

  Hereinafter, the operation of the conventional LPC encoder configured as described above will be described.

First, the correlator 10 autocorrelation of the input signal x [n], LP coefficient calculator 11, LP coefficient _{a n} and a gain by processing the autocorrelation value _r x calculated by the correlator 10 [n] G is calculated. Here, the V / UV determination unit 12 determines whether the input signal x [n] is a voice (V) signal or a non-voice (UV) signal, and determines the V / UV instruction bit and the voice (V). The signal is output, and the pitch calculation unit 13 calculates the pitch P of the voice (V) signal output from the V / UV determination unit 12.

Accordingly, when the V / UV instruction bit indicates a voice (V) signal, the parameter coding unit 14 performs parameter coding on the LP coefficient a _n , the gain G, and the pitch P received from the LP coefficient calculation unit 11 and the pitch calculation unit 13 ( Encode at a low bit rate) and output a bitstream. Thereafter, a control unit (not shown) processes the bit stream and outputs the bit stream to a radio unit (not shown), and the radio unit converts the signal output from the control unit into a radio signal and transmits the radio signal. .

  In this way, in order to transmit the audio signal at a low bit rate in the mobile communication terminal, the conventional technique performs LPC coding. However, since conventional LPC coding normally uses LP coefficients, human auditory characteristics are not considered. Therefore, the conventional LPC coding that operates at a low bit rate has the disadvantages that the compression efficiency is low (1200 kbps to 2400 kbps) and the sound quality is poor.

  An object of the present invention is to provide a voice coding apparatus and method for a mobile communication terminal that improves compression efficiency and sound quality by performing LPC coding using PLP coefficients.

  In order to achieve such an object, an LPC encoder of a mobile communication terminal according to the present invention includes a PLP coefficient calculation unit that processes an input signal to calculate a PLP coefficient and a gain, and the input signal is an audio signal. Or a non-audio signal, and if the input signal is an audio signal, a V / UV determination unit that outputs the determination signal and the corresponding audio signal, and an input output from the V / UV determination unit A pitch calculating unit that calculates a pitch of a signal, and a parameter coding unit that performs low bit rate coding using the PLP coefficient, gain, and pitch based on the determination signal.

  In order to achieve the above object, a low bit rate speech coding method for a mobile communication terminal according to the present invention includes a step of processing an input signal to calculate a PLP coefficient and a gain, and the input signal is a speech signal. Determining whether the input signal is an audio signal, and outputting the determination signal and the audio signal; and determining a pitch of the input signal output from the V / UV determination unit. Calculating, and performing low bit rate coding using the PLP coefficient, gain, and pitch based on the decision signal.

  Preferably, the audio signal is a speech signal.

  Preferably, the order of the PLP coefficient is about the 7th order in the case of an 8 kHz sampling rate.

In order to achieve the above object, the present invention provides, for example, the following means.
(Item 1)
In mobile communication terminals,
A PLP coefficient calculation unit that processes an input signal and calculates a PLP (Perceptual Linear Prediction) coefficient and a gain;
A V / UV determination unit that determines whether the input signal is a voice (Voiced) signal or a non-voice (Unvoiced) signal, and outputs the determination signal and the corresponding voice signal if the input signal is a voice signal When,
A pitch calculation unit that calculates the pitch of the input signal output from the V / UV determination unit;
And a parameter coding unit that performs low bit rate coding using the PLP coefficient, gain, and pitch based on the determination signal.
(Item 2)
The speech coding apparatus according to item 1, wherein the speech signal is a speech signal.
(Item 3)
The speech coding apparatus according to item 1, wherein the determination signal is a bit value indicating whether the input signal is a speech signal or a non-speech signal.
(Item 4)
The speech coding apparatus according to item 1, wherein the order of the PLP coefficient is about 7th in the case of an 8 kHz sampling rate.
(Item 5)
Processing the input signal to calculate PLP coefficients and gains;
Determining whether the input signal is an audio signal or a non-audio signal, and outputting the determination signal and the audio signal when the input signal is an audio signal;
Calculating the pitch of the input signal output from the V / UV determining unit;
And performing low bit rate coding using the PLP coefficient, gain, and pitch based on the determination signal.
(Item 6)
The method of claim 5, wherein the voice signal is a speech signal.
(Item 7)
The step of calculating the PLP coefficient and gain includes:
Fast Fourier transforming the input signal;
Removing noise components in units of frequencies by performing integration and resampling on the fast Fourier transformed speech signal;
The speech signal from which the noise component has been removed is subjected to equalizing processing and loudness correction to a sound component having a magnitude suitable for human hearing, and then matched to an appropriate power.
Obtaining a linear equation system by performing inverse discrete Fourier transform on the power-matched speech signal;
6. The voice coding method of a mobile communication terminal according to item 5, comprising a step of obtaining a PLP coefficient and a gain by performing a cepstral circulation process on the linear equation set.
(Item 8)
The method of claim 5, wherein the order of the PLP coefficient is about 7th in the case of an 8 kHz sampling rate.

  The present invention has an effect of improving the compression rate by performing LPC using PLP coefficients and transmitting an audio signal at a more efficient low bit rate.

  In addition, the present invention has an effect that by using the PLP coefficient as a parameter, a higher compression rate can be realized and higher signal quality can be expected than when the existing LP coefficient is used.

  Therefore, the present invention can be used for speech coding and decoding using a low bit rate, or for an apparatus that performs speech synthesis in a small space using PLP parameters.

  In addition, the present invention does not require high sound quality, but can be used for speech coding for applications that are sufficiently audible. In addition, the present invention is effective for storing data at a high compression rate in an embedded system having a limited memory, and for voice calls on the Internet or the like that require a low bit rate.

  Hereinafter, preferred embodiments of the present invention will be described with reference to the drawings.

  The present invention provides low bit rate speech coding using PLP that can perform lower-order coding than LPC in order to perform speech coding having a high compression rate.

  First, the differences between PLP and LP are as follows.

Since the LP is known, the formula for obtaining it is not described. The LP basically obtains an LPE _ak such that the value of e [n] is minimized by MSE (mean squared error) according to the following equation (1).

前記求められたＬＰ係数ａ_ｋの次数は、８ｋＨｚサンプリングレートの場合、８〜１２次程度となる。従って、前記求められたＬＰ係数ａ_ｋがＬＰを利用する各種コーディング（ＬＰＣ、ＣＥＬＰ、ＭＥＬＰ、ＲＥＬＰ）に使用される。このような内容は、『Ｗ．Ｂ．Ｋｌｅｉｊｉｎ， ａｎｄ Ｋ．Ｋ．Ｐａｌｉｗａｌ， Ｓｐｅｅｃｈ ｃｏｄｉｎｇ ａｎｄ ｓｙｎｔｈｅｓｉｓ， Ａｍｓｔｅｒｄａｍ， ｔｈｅ Ｎｅｔｈｅｒｌａｎｄｓ： Ｅｌｓｅｖｉｅｒ， １９９５』に詳細に記載されている。

The order of the obtained LP coefficient _ak is about 8 to 12 in the case of an 8 kHz sampling rate. Therefore, the obtained LP coefficient _ak is used for various types of coding (LPC, CELP, MELP, RELP) using LP. Such contents are “W. B. Kleijin, and K.K. K. Paliwal, Speech coding and synthesis, Amsterdam, the Netherlands: Elsevier, 1995.

  The PLP first appeared in a Hermansky paper in 1990, and uses human auditory characteristics in the same way as the existing MFCC (Mel-Frequency Cepstial Coefficient). Accordingly, when performing LPC for a low bit rate, the present invention performs low bit rate speech coding (LPC) using PLP coefficients instead of LP coefficients.

  That is, the present invention obtains a spectrum using the PLP coefficient. Since the PLP coefficient reflects human auditory characteristics, the spectrum obtained using this PLP coefficient has a larger error than LP as an MSE concept. Becomes smaller. In the case of transmission of LP coefficients, transmission is generally performed at the 10th order at the 8 kHz sampling rate. However, in the case of transmission of PLP coefficients, transmission is performed at the seventh order, so that the bit rate can be lowered.

  FIG. 1 is a configuration diagram of an LPC encoder using a PLP coefficient according to the present invention.

  Referring to FIG. 1, the LPC encoder using the PLP coefficient is the same as the conventional LPC encoder of FIG. 3 except that the correlator 10 is removed and the LP coefficient calculator 11 is replaced with a PLP coefficient calculator 20. It is the same.

PLP coefficient calculator 20 calculates the PLP coefficient auditory characteristics is considered by processing the speech signal S [n] _{a P} and a gain G.

  Hereinafter, the operation of the LPC encoder using another PLP coefficient according to the present invention configured as described above will be described with reference to the drawings.

  First, the PLP coefficient calculation unit 20 receives the speech signal S [n] and sequentially performs the operations shown in FIG. 2 to calculate the PLP coefficient aP and the gain G.

  That is, the PLP coefficient calculation unit 20 first performs fast Fourier transform (FFT) on the input signal that is the speech signal S [n], and integrates and resampling the fast Fourier transformed speech signal. By performing the above, noise components are removed from the speech signal S [n] in frequency units.

  When the noise component is removed, the PLP coefficient calculation unit 20 performs equalizing processing and loudness correction of the Fourier-transformed speech signal to a sound component having a size suitable for human hearing, and then performs human loudness correction. Match power suitable for listening.

When the power matching is completed, the PLP coefficient calculator 20 performs inverse discrete Fourier transform on the speech signal and then obtains a linear equation system (Set of Linear equations) from the speech signal. Accordingly, the PLP coefficient calculation unit 20 outputs cepstral coefficients of the PLP model, that is, the PLP coefficient a _P , by performing cepstral circulation processing on the linear equation system. That, PLP coefficient calculator 20 outputs to the parameter coding unit 23 a PLP coefficient of low reflecting the human auditory characteristics orders a _P and a gain G as parameters values.

  Here, the V / UV determination unit 21 outputs the V / UV instruction bit and transmits the speech signal S [n] to the pitch calculation unit 22, and the pitch calculation unit 22 determines the speech signal S [n]. The pitch P is calculated.

Accordingly, the parameter coding unit 23 codes (encodes at a low bit rate) the V / UV instruction bit value, the PLP coefficient a _P , the gain G, and the pitch P received from the PLP coefficient calculation unit 20 and the pitch calculation unit 22 to generate bits. Output a stream. Preferably, the order of the PLP coefficient _{a P} is the transmission, in the case of 8kHz sampling rate, the 7 th order. Thereafter, the control unit (not shown) processes the bit stream and outputs it to a radio unit (not shown), and the radio unit converts the signal output from the control unit into a radio signal and transmits it.

  As mentioned above, although this invention has been illustrated using preferable embodiment of this invention, this invention should not be limited and limited to this embodiment. It is understood that the scope of the present invention should be construed only by the claims. It is understood that those skilled in the art can implement an equivalent range based on the description of the present invention and the common general technical knowledge from the description of specific preferred embodiments of the present invention.

  By performing LPC (Linear Predictive Coding) coding using PLP (Perceptual Linear Prediction) coefficients, it is possible to realize a higher compression ratio than when using existing LP coefficients, and the voice of a mobile communication terminal that can guarantee high sound quality Provide a coding method.

  A voice coding apparatus of a mobile communication terminal includes a PLP coefficient calculation unit 20 that processes an input signal to calculate a PLP coefficient and a gain, and the input signal is a voice (Voiced) signal or a non-voiced (Unvoiced) signal. If the input signal is an audio signal, the V / UV determination unit 21 that outputs the determination signal and the corresponding audio signal, and the pitch of the input signal output from the V / UV determination unit 21 are calculated. A pitch calculation unit 22 and a parameter coding unit 23 that performs low bit rate coding using the PLP coefficient, gain, and pitch based on the determination signal are included.

It is a figure which shows the structure of the LPC encoder using the PLP coefficient based on this invention. It is a figure for demonstrating in detail the step which calculates the PLP coefficient in FIG. It is a figure which shows the structure of the LPC encoder using the conventional LP coefficient.

Explanation of symbols

20: PLP coefficient calculation unit 21: V / UV determination unit 22: Pitch calculation unit 23: Parameter coding unit

Claims (8)

In mobile communication terminals,
A PLP coefficient calculation unit that processes an input signal and calculates a PLP (Perceptual Linear Prediction) coefficient and a gain;
A V / UV determination unit that determines whether the input signal is a voice (Voiced) signal or a non-voice (Unvoiced) signal, and outputs the determination signal and the corresponding voice signal when the input signal is a voice signal. When,
A pitch calculation unit that calculates the pitch of the input signal output from the V / UV determination unit;
And a parameter coding unit that performs low bit rate coding using the PLP coefficient, gain, and pitch based on the determination signal.   The speech coding apparatus according to claim 1, wherein the speech signal is a speech signal.   The speech coding apparatus according to claim 1, wherein the determination signal is a bit value indicating whether the input signal is a speech signal or a non-speech signal.   The speech coding apparatus according to claim 1, wherein the order of the PLP coefficient is about 7th in the case of an 8 kHz sampling rate. Processing the input signal to calculate PLP coefficients and gains;
Determining whether the input signal is an audio signal or a non-audio signal, and outputting the determination signal and the audio signal when the input signal is an audio signal;
Calculating the pitch of the input signal output from the V / UV determining unit;
And performing low bit rate coding using the PLP coefficient, gain, and pitch based on the determination signal.   The method of claim 5, wherein the voice signal is a speech signal. The step of calculating the PLP coefficient and gain includes:
Fast Fourier transforming the input signal;
Removing noise components in units of frequencies by performing integration and resampling on the fast Fourier transformed speech signal;
The speech signal from which the noise component has been removed is subjected to equalizing processing and loudness correction to a sound component having a magnitude suitable for human hearing, and then matched to an appropriate power.
Obtaining a linear equation system by performing inverse discrete Fourier transform on the power-matched speech signal;
6. The voice coding method of a mobile communication terminal according to claim 5, comprising a step of obtaining a PLP coefficient and a gain by performing a cepstral circulation process on the linear equation set. .   The method of claim 5, wherein the order of the PLP coefficient is about 7th in the case of an 8 kHz sampling rate.

Images