KR100911994B1 - Method and apparatus for encoding/decoding signal having strong non-stationary properties using hilbert-huang transform - Google Patents

Abstract

The present invention relates to coding and decoding of speech and audio signals, and more particularly, to simultaneously encode input signals having strong non-stationary characteristics such as speech and audio signals using HHT (Hilbert-Huang Transform). And an apparatus and method capable of decoding.

The present invention provides a method for separating input signals into signals having different frequency bands by applying HHT, independently encoding each of the separated signals, and multiplexing each independently encoded signal. It provides an encoding method that includes.

Hilbert transform, EMD, HHT, intrinsic mode function,

Description

TECHNICAL AND APPARATUS FOR ENCODING / DECODING SIGNAL HAVING STRONG NON-STATIONARY PROPERTIES USING HILBERT-HUANG TRANSFORM}

The present invention relates to coding and decoding of speech and audio signals. More particularly, the present invention relates to an input signal having a strong non-stationary characteristic such as speech and audio signals using HHT (Hilbert-Huang Transform). An apparatus and method capable of encoding and decoding at the same time.

In general, voice and speech codecs and audio and music codecs have been developed independently.

Encoders that process single content such as voice or audio are used very effectively in voice communication in mobile communication and music players such as portable MP3 players. However, in the future, more convergence of broadcasting and communication is required to transmit various contents, and also air Due to the fundamental capacity limitation of the channel, low bit rate transmission is inevitable. In order to solve this problem, a low bit rate coding technique capable of efficiently integrating speech and audio signals is required.

Conventional coding techniques have been independently developed for voice and audio, and an optimal technique called a Code Excited Linear Prediction (CELP) structure for voice signals has been developed, and AAC + technology for audio signals has been completed. However, each coding technique has a problem of causing a large performance degradation for an input signal having a different area characteristic. That is, the CELP coding method cannot represent accurate frequency information by incorrect modeling when an audio signal is input, and a very large performance degradation occurs, and when a speech signal is input to a low bit rate AAC +, due to limited bits. Poor performance due to inaccurate representation of harmonic structure.

Recently, in 3GPP, AMR-WB + having multiple structures has been standardized to process signals having various characteristics. The AMR-WB + is a dual structure having AMR-WB having a CELP structure and a TCX (Transform Coded Excitation) module having a transform structure. The AMR-WB + has a CELP and a TCX structure selected according to an input signal. In the end, it provides the basic structure of processing the voice with CELP and audio with the Transform encoder, and it has excellent performance for both voice and audio at low bit rate of 32kbps or less with one encoder. However, the biggest problem with AMR-WB + is that the performance of the TCX module is lower than that of AAC +, which results in poor performance for audio signals.

Table 1 below shows the results of officially measuring the listening performance of the two encoding techniques for speech and audio signals.

Referring to [Table 1], it is shown that the performance of AMR-WB + is excellent for voice content and AAC + is excellent for audio content at all bit rates. Therefore, the two encoding techniques have very opposite characteristics and performances, and thus, both encoding techniques cannot be optimal encoding techniques in an application field dealing with various characteristic inputs.

In other words, the integrated coding technology according to the prior art has advantages and disadvantages that AMR-WB + and AAC + are opposite to each other, and both coding technologies have not yet provided the best performance. Thus, there is a need for an integrated encoding technique that provides excellent performance for speech and audio signals.

SUMMARY OF THE INVENTION The present invention has been made to solve the problems of the prior art as described above, and an object of the present invention is to provide an apparatus and method capable of efficiently encoding and decoding voice and audio signals simultaneously.

Another object of the present invention is to provide an apparatus and method capable of efficiently encoding and decoding an input signal having a strong non-normal characteristic.

In order to achieve the above object and solve the problems of the prior art, the present invention, the signal decomposition unit for decomposing the input signal into signals having different frequency bands; A phase and amplitude information extracting unit extracting instantaneous phase and instantaneous amplitude information of each of the separated signals by Hilbert transforming the decomposed signal; And an encoding unit encoding each of the decomposed signals using the extracted instantaneous phase and instantaneous amplitude information.

According to another aspect of the present invention, there is provided an encoding apparatus, including: a signal decomposition unit for decomposing an input signal into signals having different frequency bands; A first encoder to encode a signal belonging to a high frequency region among the decomposed signals by applying a CELP-based encoding or an audio signal encoding algorithm; And a second encoder for extracting respective instantaneous phase and instantaneous amplitude information by applying a Hilbert transform to the remaining signals that do not belong to the high frequency region among the decomposed signals, and encoding the extracted instantaneous phase and instantaneous amplitude information. do.

In addition, the present invention comprises the steps of decomposing the signal into a signal having a different frequency band by applying the HHT to the input signal; Independently encoding each of the decomposed signals; And multiplexing each of the independently encoded signals.

Additional features and advantages of the present invention will become more apparent from the following detailed description of the embodiments for carrying out the invention, which is to be understood by the embodiments and drawings, although the invention is not limited thereto. The scope of the present invention is not limited to these embodiments, which can be variously modified and modified by those skilled in the art. Accordingly, the invention idea should be construed only by the claims set forth below, and all equivalent or equivalent modifications thereof should be construed as falling within the scope of the invention idea.

According to the present invention, it is possible to provide an efficient method for simultaneously encoding a voice and an audio signal having strong irregularity with one integrated codec. It can be effectively applied to an encoder for voice and audio signals among multimedia encoders to be used later in mobile communication / broadcasting. Since the encoding technique according to the present invention is a new method different from the existing speech and audio encoding techniques, it is possible to save a huge amount of technical fees paid by using the existing encoding technique.

Hereinafter, an apparatus and method for encoding / decoding audio and audio signals using HHT according to an embodiment of the present invention will be described in detail with reference to the accompanying drawings.

The basic principle of the present invention is to decompose an input signal into intrinsic mode function (IMF) signals having different frequency band ranges using the Hilbert-Huang Transform (HHT) method, and the instantaneous phase and By extracting and quantizing amplitude information, signals having strong non-stationary characteristics are more efficiently encoded. The HHT is a name combined with the empirical mode decomposition (EMD) method and the Hilbert transform.

In the present invention, the encoding method of the instantaneous phase information is performed based on a polynomial phase modeling method in an arbitrary short time period. For efficient phase coding, the phase information is modeled polynomial in any short time interval, and then the polynomial coefficients are quantized. In consideration of the fact that the frequency change with time in the short interval decreases as the index of the extracted IMF signal increases, such a phase coding scheme is very efficient as the index of the IMF signal increases. For reference, the phase is obtained as a cumulative sum of frequency components over time.

The index of the IMF signal refers to a sequence of signals decomposed from the high frequency band to the low frequency band by the HHT method.

In the present invention, the instantaneous amplitude information is encoded based on the vector quantization method. The instantaneous amplitude information represents the overall envelope of a signal having a positive value, and since they have a relatively high correlation with each other in the time domain, they can be efficiently encoded even by a vector quantization method.

1 is a block diagram showing a schematic configuration of an encoding apparatus according to the present invention.

Referring to FIG. 1, the encoding apparatus may include a signal decomposition unit 110 that decomposes an input signal into signals having different frequency bands, and Hilbert transforms the decomposed signal to generate instantaneous phase and instantaneous amplitude of each of the decomposed signals. And a coding unit 130 for encoding each of the decomposed signals using the extracted phase and amplitude information extracting unit 120 and the extracted instantaneous phase and instantaneous amplitude information.

The signal decomposition unit 110 decomposes the input signal by applying EMD (empirical mode decomposition) to the input signal. The input signal is mainly a signal having a strong non-normal characteristic. For example, the input signal may be a signal in which a voice and an audio signal are mixed or one of a voice or an audio signal.

The phase and amplitude information extractor 120 corresponds to a Hilbert transformer for performing Hilbert transform.

The encoder 130 includes a plurality of encoders for independently encoding each of the decomposed input signals.

2 is a block diagram showing a detailed configuration of an encoding apparatus according to the present invention.

Referring to FIG. 2, the encoding apparatus according to the present invention includes a signal decomposition unit 210 for applying an EMD to an input signal and decomposing the signal into IMF signals having different frequency bands for each frame of the input signal, and the IMF signal. A plurality of Hilbert transform units 220 for extracting respective phase and amplitude information, and a plurality of encoders 230 (encoders 1 to encoder N) encoded using a phase and amplitude algorithm of each IMF signal. And a multiplexing unit 250 for multiplexing (bit packing) each of the encoded IMF signals.

3 is a block diagram illustrating a detailed configuration of an encoding unit encoding phase and amplitude for each IMF signal according to an embodiment of the present invention.

Referring to FIG. 3, the encoder 230 illustrated in FIG. 2 includes a phase encoder 330 and an amplitude encoder 340, respectively.

는 i번째 IMF 신호에 대한 프레임내의 순간 위상 벡터를 의미하고,

는 순간 진폭 벡터를 나타낸다. 그리고

는 순간 위상 벡터를 다항식(polynomial)으로 모델링하였을 때의 계수벡터를 나타낸다.In FIG. 3,

Is the instantaneous phase vector in the frame for the i th IMF signal,

Denotes the instantaneous amplitude vector. And

Denotes a coefficient vector when the instantaneous phase vector is modeled as a polynomial.

The phase encoder 330 is a phase continuator 331 for performing a phase continuity process on the instantaneous phase information extracted by the Hilbert transform unit 320 and a matrix form of the instantaneous phase information that has been subjected to the continuity process. A polynomial modeling unit 332 for performing linear modeling and calculating coefficients of the polynomial using a least square method from the linear model of the matrix form and a quantization unit 333 for quantizing the coefficients of the calculated polynomial. .

The phase continuity unit 331 performs phase unwrapping on the extracted phase information.

을 p차의 다항식으로 모델링한다.The polynomial modeling unit 332 is instantaneous phase information at time n, as shown in Equation 1 below.

Is modeled as a p-order polynomial.

Next, the polynomial modeling unit 332 is a linear matrix having a length vector of N length and a polynomial coefficient vector of length p + 1, as shown in Equation 1 below. Transform into a model.

From the linear model of Equation 2, polynomial coefficients can be obtained as shown in Equation 3 by a least square method.

[Equation 3]

의 연산은 매트릭스(matrix)의 역변환 과정 때문에 복잡하지만, 길이 N과 다항식의 차수 p가 정해지면 X 매트릭스의 성분들도 상수값으로 고정되기 때문에, 이를 미리 계산하여 테이블로 사용함으로써 연산 효율을 높일 수 있다. 또한, 각 IMF 신호의 다른 위상 변화 특성을 고려하여, IMF 신호별로 각기 다른 모델링 길이 N과 다항식 차수 p를 적용하고 양자화 비트수도 다르게 사용하는 것도 가능하다.Substantially, in [Equation 3],

The operation of is complicated by the inverse transformation of the matrix, but since the components of the X matrix are fixed to constant values when the length N and the degree p of the polynomial are determined, the computational efficiency can be increased by using them as a table in advance. have. In addition, in consideration of different phase change characteristics of each IMF signal, it is also possible to apply different modeling lengths N and polynomial order p for each IMF signal and use different numbers of quantization bits.

The quantization unit 333 performs split scalar quantization of polynomial coefficients extracted by modeling an instantaneous phase vector. The split scalar quantization is performed to minimize the mean square error in the synthesized signal region defined by Equation 4 below.

은 양자화되지 않은 i 번째 IMF 신호까지 더한 목적(target) 신호이고,

은 i-1 번째까지 양자화 과정을 통해 구한 신호 에 i 번째 IMF 신호의 위상 정보를 양자화하여 더한 신호를 의미한다.However, since all coefficients of the polynomial are required to synthesize the actual signal, a ₀ to a _{p -1} are quantized in the coefficient domain instead of the synthesized signal domain, and in the process of quantizing a _p [Equation 4] The method of minimizing the mean square error is applied. In [Equation 4] above

Is the target signal plus the i-th unquantized iMF signal,

Denotes a signal obtained by quantizing the i-1 th signal and quantizing phase information of the i th IMF signal.

The amplitude encoder 340 performs vector quantization in the amplitude signal region, calculates frame energy in order to increase quantization efficiency, and normalizes the signal using the result, and then quantizes it.

The amplitude encoder 340 is an energy calculator 341 that calculates the frame energy of the extracted instantaneous amplitude, a vector quantizer 344 and 345 which normalizes the calculated frame energy value to vector quantize, and the extraction. And scalar quantization units 342 and 343 for scalar quantizing the instantaneous amplitudes in the logarithmic region.

4 is a block diagram showing a schematic configuration of a decoding apparatus according to the present invention.

Referring to FIG. 4, the decoding apparatus according to the present invention receives a demultiplexer 410 for receiving and demultiplexing a bit stream encoded for each signal of a different frequency band, and independently decodes each of the demultiplexed signals. A plurality of decoding units 420 (decoding unit 1 to decoding unit N) to perform and a summation unit 430 for restoring the original signal by summing the output signal of each of the decoding unit.

FIG. 5 is a block diagram showing the detailed configuration of the decoding unit shown in FIG. 4.

Referring to FIG. 5, the decoder 420 illustrated in FIG. 4 includes phase decoders 421, 422, and 423, amplitude decoders 424, 425, 426, and 427, and a multiplier 428, respectively. It is configured by.

를 얻어내고, 이를 바탕으로 역다항식 모델링부(PPM-1)(422)에 의하여 하기 [수학식 5]와 같은 위상 벡터

를 추출한다.The phase decoders 421, 422, and 423 are polynomial coefficient vectors modeling instantaneous phase vectors through an inverse quantizer 421.

And a phase vector as shown in [Equation 5] by the inverse polynomial modeling unit (PPM-1) 422

Extract

Since the synthesized IMF signals are real numbers, the cosine function is applied to the phase information by the cosine function applying unit 423.

The amplitude decoders 424, 425, 426, 427 are performed by a normalized amplitude vector dequantizer 424 and a frame energy dequantizer 423.

를 출력한다.The amplitude vector inverse quantizer 424 is a standardized amplitude vector

Outputs

와 곱해진다. The output value of the frame energy dequantization unit 423 passes through an inverse exponential function application unit 426 to restore the quantized frame energy in the logarithmic region, and is decoded by the multiplier 427.

Multiplied by

As a result, the output values of the phase decoders 421, 422, and 423 and the amplitude decoders 424, 425, 426, and 427 are multiplied by the multiplier 428 to obtain an i-th intrinsic mode function (IMF) signal. Is restored to the form as shown in [Equation 6].

6 is a block diagram showing a configuration of an encoding apparatus according to another embodiment of the present invention.

Referring to FIG. 6, an encoding apparatus according to another embodiment of the present invention is

A signal decomposition unit (EMD) for decomposing an input signal into signals having different frequency bands, and a method for encoding by applying a CELP-based encoding or an audio signal encoding algorithm to signals belonging to a high frequency region among the decomposed signals. The first encoder 620 and the Hilbert transform are applied to the remaining signals that do not belong to the high frequency region of the decomposed signals to extract respective instantaneous phase and instantaneous amplitude information and encode the extracted instantaneous phase and instantaneous amplitude information. It is configured to include a second encoder 630.

That is, another embodiment of the present invention shows an embodiment of a hybrid encoding algorithm using a mixture of an IMF signal encoding method in a Hilbert region and a CELP-based encoding or an existing audio encoding scheme in a time domain. . Indeed, since the first or second IMF signals mostly exist in the high frequency region and the change of the instantaneous phase is large, the instantaneous phase modeling efficiency using the polynomial is inferior. For these signals, coding efficiency can be improved by using a code-excited linear prediction (CELP) based method or an audio coding algorithm used in conventional speech coding.

In FIG. 6, the psychoacoustic modeling unit 610 applies a psychoacoustic model (PAM) to the input signal to remove unnecessary components that are not acoustically important from the input signal.

In addition, the psychoacoustic model may be applied to an input signal or may be applied after Hilbert transform of each IMF signal, as shown in FIG. 6, thereby increasing coding efficiency.

The encoding and decoding method according to the present invention can be implemented in the form of program instructions that can be executed by various computer means and recorded in a computer readable medium. The computer readable medium may include program instructions, data files, data structures, etc. alone or in combination. Program instructions recorded on the media may be those specially designed and constructed for the purposes of the present invention, or they may be of the kind well-known and available to those having skill in the computer software arts. Examples of computer-readable recording media include magnetic media such as hard disks, floppy disks, and magnetic tape, optical media such as CD-ROMs, DVDs, and magnetic disks, such as floppy disks. Magneto-optical media, and hardware devices specifically configured to store and execute program instructions, such as ROM, RAM, flash memory, and the like. The medium may be a transmission medium such as an optical or metal wire, a waveguide, or the like including a carrier wave for transmitting a signal specifying a program command, a data structure, or the like. Examples of program instructions include not only machine code generated by a compiler, but also high-level language code that can be executed by a computer using an interpreter or the like. The hardware device described above may be configured to operate as one or more software modules to perform the operations of the present invention, and vice versa.

As described above, the present invention has been described by way of limited embodiments and drawings, but the present invention is not limited to the above embodiments, and those skilled in the art to which the present invention pertains various modifications and variations from such descriptions. This is possible.

Therefore, the scope of the present invention should not be limited to the described embodiments, but should be determined not only by the claims below but also by the equivalents of the claims.

1 is a block diagram showing a schematic configuration of an encoding apparatus according to the present invention.

2 is a block diagram showing a detailed configuration of an encoding apparatus according to the present invention.

3 is a block diagram illustrating a detailed configuration of an encoding unit encoding phase and amplitude for each IMF signal according to an embodiment of the present invention.

4 is a block diagram showing a schematic configuration of a decoding apparatus according to the present invention.

FIG. 5 is a block diagram showing the detailed configuration of the decoding unit shown in FIG. 4.

6 is a block diagram showing a configuration of an encoding apparatus according to another embodiment of the present invention.

Claims (14)

A signal decomposition unit for decomposing the input signal into signals having different frequency bands; A phase and amplitude information extracting unit extracting instantaneous phase and instantaneous amplitude information of each of the separated signals by Hilbert transforming the decomposed signal; And And an encoder which encodes each of the decomposed signals using the extracted instantaneous phase and instantaneous amplitude information. The method of claim 1, wherein the encoding unit, A phase encoder for encoding the coefficients of the polynomial by modeling the extracted instantaneous phase information into a polynomial; And And an amplitude encoder configured to vector-quantize and extract the extracted instantaneous amplitude information. The method of claim 2, wherein the phase encoding unit, A phase continuity unit configured to perform a phase continuation process on the extracted instantaneous phase information; A phase modeling unit performing linear modeling of matrix type on the instantaneous phase information subjected to the sequencing process and calculating coefficients of the polynomial using a least square method from the matrix type linear model; And And a quantization unit configured to quantize the calculated coefficients of the polynomial. The method of claim 3, wherein the phase modeling unit, And encoding coefficients of the polynomial using a pre-stored transform matrix according to a modeling length and a polynomial order. The method of claim 2, wherein the amplitude encoder, An energy calculator configured to calculate energy of the extracted instantaneous amplitude; A vector quantizer which normalizes the calculated energy value and quantizes the vector; And And a scalar quantizer for scalar quantizing the extracted instantaneous amplitude in a logarithmic region. A signal decomposition unit for decomposing the input signal into signals having different frequency bands; A first encoder to encode a signal belonging to a high frequency region among the decomposed signals by applying a CELP-based encoding or audio signal encoding algorithm; And A second encoder which extracts respective instantaneous phase and instantaneous amplitude information by applying a Hilbert transform to the remaining signals that do not belong to the high frequency region among the decomposed signals, and encodes the extracted instantaneous phase and instantaneous amplitude information. Encoding device. The encoding apparatus of claim 6, further comprising a psychoacoustic modeling unit applying the psychoacoustic model to the input signal and providing the psychoacoustic model to the signal separation unit. The encoding apparatus of claim 6, wherein the second encoder removes unnecessary components by applying a psychoacoustic model to the signal to which the Hilbert transform is applied. A demultiplexer configured to receive and demultiplex the encoded bit streams for signals of different frequency bands;  A phase inverse quantizer for extracting phase information from a polynomial coefficient vector modeling a phase vector by inverse quantization of each demultiplexed signal; An amplitude dequantization unit configured to dequantize each of the demultiplexed signals and extract amplitude information from an energy of an amplitude vector and an amplitude vector; A multiplier for multiplying the extracted phase information and amplitude information to restore signals of different frequency bands; And And a summation unit configured to sum the restored signals of different frequency bands and restore the original signal. The method of claim 9, wherein the original signal, A decoding apparatus, characterized in that a mixture of speech and audio signals or one of speech and audio signals. Applying HHT to the input signal and decomposing the signal into signals having different frequency bands; Independently encoding each of the decomposed signals; And Multiplexing the independently encoded respective signals. The method of claim 11, wherein the independently encoding each of the decomposed signals comprises: Extracting instantaneous phase information and instantaneous amplitude information for each of the decomposed signals; And encoding the instantaneous phase and amplitude information by applying different encoding methods according to the extracted characteristics of the extracted instantaneous phase and amplitude information. The method of claim 12, wherein at least one of the different encoding methods, And encoding the coefficients of the polynomial by modeling the extracted instantaneous phase information into a polynomial and performing vector quantization on the extracted instantaneous amplitude information. Receiving and demultiplexing a bit stream encoded for signals of different frequency bands; Extracting phase information from a polynomial coefficient vector modeling a phase vector of each of the demultiplexed signals, and extracting amplitude information from energy of amplitude vectors and amplitude vectors of each of the demultiplexed signals; Multiplying the extracted phase information and amplitude information of each of the demultiplexed signals; And And summing respective signals obtained by multiplying the extracted phase information and amplitude information.

Images