KR19990054249A - ADP-CM method using Walsh-Hadamard transformation - Google Patents
ADP-CM method using Walsh-Hadamard transformation Download PDFInfo
- Publication number
- KR19990054249A KR19990054249A KR1019970074054A KR19970074054A KR19990054249A KR 19990054249 A KR19990054249 A KR 19990054249A KR 1019970074054 A KR1019970074054 A KR 1019970074054A KR 19970074054 A KR19970074054 A KR 19970074054A KR 19990054249 A KR19990054249 A KR 19990054249A
- Authority
- KR
- South Korea
- Prior art keywords
- adaptive
- signal
- error signal
- quantized
- wht
- Prior art date
Links
Landscapes
- Compression, Expansion, Code Conversion, And Decoders (AREA)
Abstract
본 발명은 ADPCM을 위한 적응 예측기에 입력되는 양자화된 오차신호를 왈시-하다마드 변환(WHT)에 의해 직교변환하고 전력 정규화함으로써 계산량을 줄일 수 있고 수렴 성능을 향상시킬 수 있도록 한 WHT를 이용한 ADPCM 방법에 관한 것으로, 입력되는 PCM 데이터의 포맷을 변환한 후 출력신호와의 차신호를 구하고, 이어 적응 양자화기를 통해 상기 차신호를 양자화하고, 다시 역 적응 양자화기를 통해 상기 양자화된 차신호를 역 양자화하여 오차신호를 구하며, 이후 WHT에 의해 상기 양자화된 오차신호를 직교변환하고, 전력 정규화 과정을 수행하여 적응 예측기를 통해 적응 예측하도록 함으로써, WHT에 의해 계산량을 줄일 수 있음은 물론 직교변환을 통해 입력신호의 인접 샘플간의 상관도를 줄일 수 있어 ADPCM을 위한 적응 예측기의 수렴 성능을 향상시킬 수 있게 되는 효과가 있다.According to the present invention, the quantized error signal input to the adaptive predictor for ADPCM is orthogonally transformed by the Walsh-Hadamard transform (WHT) and the power is normalized to reduce the calculation amount and improve the convergence performance. And converting the format of the input PCM data to obtain a difference signal with the output signal, and then quantizing the difference signal through an adaptive quantizer, and inversely quantizing the quantized difference signal through an inverse adaptive quantizer. An error signal is obtained, and then the orthogonal transform of the quantized error signal is performed by the WHT, and a power normalization process is performed to make an adaptive prediction through an adaptive predictor. It is possible to reduce the correlation between neighboring samples of P and improve the convergence performance of the adaptive predictor for ADPCM. It is effective to be.
Description
본 발명은 에이디피씨엠(Adaptive Differential Pulse Code Modulation ; 이하, 'ADPCM'이라 칭함)에 있어서, 적응 예측기에 입력되는 양자화된 오차신호를 왈시-하다마드 변환(Walsh-Hadamard Transform ; 이하, 'WHT'라 칭함)에 의해 직교변환하고 전력 정규화함으로써 계산량을 줄일 수 있고 수렴 성능을 향상시킬 수 있도록 한 WHT를 이용한 ADPCM 방법에 관한 것이다.According to the present invention, a Walsh-Hadamard Transform (WHT) is used to describe a quantized error signal input to an adaptive predictor in Adaptive Differential Pulse Code Modulation (hereinafter, referred to as 'ADPCM'). By the orthogonal transformation and power normalization, the present invention relates to an ADPCM method using WHT, which can reduce the calculation amount and improve the convergence performance.
일반적으로 파형 부호화의 일종인 ADPCM에서는 계산량이 적고 구현이 용이한 LMS 알고리즘을 이용한다.In general, ADPCM, which is a type of waveform encoding, uses an LMS algorithm that is low in computation and easy to implement.
즉, 도 1에 도시된 바와 같이 종래의 ADPCM 방법은, 비교부(10)에서 입력신호 s(k)와 출력신호 se(k)와의 차신호 d(k)를 구하고, 이어 적응 양자화기(Adaptive Quantizer)(20)를 통해 상기 차신호 d(k)를 일정 비트로 양자화하며, 다시 역 적응 양자화기(Inverse Adaptive Quantizer)(30)를 통해 상기 적응 양자화기(20)에서 출력된 양자화 신호 I(k)의 오차신호 dq(k)를 생성한 후, 적응 예측기(Adaptive Predictor)(40)를 통해 상기 오차신호 dq(k)를 그대로 적응 예측하여 출력신호 se(k)를 구한다.That is, as shown in FIG. 1, in the conventional ADPCM method, the comparison unit 10 obtains a difference signal d (k) between an input signal s (k) and an output signal s e (k), and then uses an adaptive quantizer ( Quantize the difference signal d (k) into a predetermined bit through an adaptive quantizer 20, and output the quantized signal I (outputted from the adaptive quantizer 20 through an inverse adaptive quantizer 30). After generating the error signal d q (k) of k), an adaptive predictor 40 adaptively predicts the error signal d q (k) to obtain an output signal s e (k).
그러나, ADPCM에 있어서, 상기와 같이 양자화된 오차신호 dq(k)를 그대로 적응 예측하여 출력신호 se(k)를 구하는 LMS 알고리즘은 적응 예측기(40)에 입력되는 신호의 자기상관행렬(auto-correlation matrix)의 고유치 퍼짐 정도(eigen value spread)가 큰 경우 수렴이 늦어지는 문제점이 있었다.However, in the ADPCM, the LMS algorithm which adaptively predicts the quantized error signal d q (k) as described above and obtains the output signal s e (k) is an autocorrelation matrix of a signal input to the adaptive predictor 40. If the eigen value spread of the -correlation matrix is large, convergence is slow.
이에 따라, RLS 알고리즘 또는 트랜스폼 도메인(Transform domain) LMS 알고리즘과 같은 여러 수렴방법을 이용하고 있으나, 이 방법 역시 상기 LMS 알고리즘에 비해 계수 갱신 횟수당 요구되는 계산량이 많아지는 단점이 있었다.Accordingly, although various convergence methods such as the RLS algorithm or the Transform domain LMS algorithm are used, this method also has a disadvantage in that the amount of calculation required per coefficient update count is larger than that of the LMS algorithm.
본 발명은 상기와 같은 문제점을 해결하기 위해 안출한 것으로서, 그 목적은 적응 예측기에 입력되는 양자화된 오차신호를 WHT에 의해 직교변환하고 전력 정규화함으로써 계산량을 줄일 수 있고 입력신호의 인접 샘플간의 상관도를 줄여 수렴 성능을 향상시킬 수 있도록 한 WHT를 이용한 ADPCM 방법을 제공하는 데에 있다.SUMMARY OF THE INVENTION The present invention has been made to solve the above problems, and its object is to orthogonally transform the quantized error signal input to the adaptive predictor by WHT and normalize the power, thereby reducing the amount of computation and the correlation between adjacent samples of the input signal. The purpose of the present invention is to provide an ADPCM method using WHT, which can improve the convergence performance by reducing the frequency.
이러한 목적을 달성하기 위한 본 발명의 WHT를 이용한 ADPCM 방법은, 입력되는 PCM 데이터의 포맷을 변환한 후 출력신호와의 차신호를 구하고, 이어 적응 양자화기를 통해 상기 차신호를 양자화하고, 다시 역 적응 양자화기를 통해 상기 양자화된 차신호를 역 양자화하여 오차신호를 구하며, 이후 WHT에 의해 상기 양자화된 오차신호를 직교변환하고, 전력 정규화(power normalization) 과정을 수행하여 적응 예측기를 통해 적응 예측하도록 함을 특징으로 한다.ADPCM method using the WHT of the present invention for achieving this object, after converting the format of the input PCM data to obtain the difference signal with the output signal, and then quantized the difference signal through an adaptive quantizer, and then back-adaptive Inverse quantization of the quantized difference signal through a quantizer obtains an error signal, and then orthogonally transforms the quantized error signal by WHT, and performs a power normalization process to perform adaptive prediction through an adaptive predictor. It features.
도 1은 종래 LMS 방법을 이용하는 에이디피씨엠 구조를 보인 도면,1 is a diagram showing an ADP system using a conventional LMS method;
도 2는 본 발명에 의한 왈시-하다마드 변환(WHT)을 이용한 에이디피씨엠 방법을 보인 흐름도.Figure 2 is a flow chart showing an ADP system using the Walsh-Hadamard transformation (WHT) according to the present invention.
<도면의 주요부분에 대한 부호의 설명><Description of the symbols for the main parts of the drawings>
10 : 비교부 20 : 적응 양자화기10: comparison unit 20: adaptive quantizer
30 : 역 적응 양자화기 40 : 적응 예측기30: Inverse Adaptive Quantizer 40: Adaptive Predictor
이하, 첨부된 도면 2를 참고하여 본 발명에 의한 WHT를 이용한 ADPCM 방법을 상세히 설명한다.Hereinafter, an ADPCM method using WHT according to the present invention will be described in detail with reference to the accompanying drawings.
먼저, ADPCM을 위해 A-law 및 μ-law로 입력되는 PCM 데이터를 유니폼(uniform)한 PCM 데이터로 변환한다(S1).First, PCM data input in A-law and μ-law for ADPCM are converted into uniform PCM data (S1).
이어, 상기 단계(S1)에서 변환된 PCM 데이터, 즉 입력신호 s(k)와 적응 예측기의 출력신호 se(k)와의 차신호 d(k)를 구하고(S2), 적응 양자화기를 통해 상기 차신호 d(k)를 일정 비트(4 비트)로 양자화한다(S3).Subsequently, the difference signal d (k) between the PCM data converted in step S1, that is, the input signal s (k) and the output signal s e (k) of the adaptive predictor is obtained (S2), and the difference is obtained through an adaptive quantizer. The signal d (k) is quantized to a predetermined bit (4 bits) (S3).
그리고, 역 적응 양자화기를 통해 상기 단계(S3)에서 양자화된 양자화 신호 I(k)를 역 양자화하여 양자화된 오차신호 dq(k)를 생성한다(S4).In operation S4, the quantized quantized signal I (k) is inversely quantized through the inverse adaptive quantizer to generate a quantized error signal d q (k).
그리고 나서, WHT에 의해 상기 양자화된 오차신호 dq(k)를 직교변환하고(S5), 전력을 정규화한다(S6).Then, orthogonal transform the quantized error signal d q (k) by WHT (S5), and normalize the power (S6).
즉, WHT를 통해 상기 양자화된 오차신호 dq(k)의 자기상관행렬의 고유치 퍼짐정도가 커지는 것을 방지하게 된다.That is, the spread of the eigenvalue of the autocorrelation matrix of the quantized error signal d q (k) is prevented from increasing through the WHT.
이때, WHT는 곱하는 값이 0과 1로 이루어져 덧셈 계산만으로 구현이 가능함에 따라 계산량을 현저히 줄일 수 있을 뿐만 아니라 직교변환의 일종으로 입력신호의 인접 샘플간의 상관도를 줄일 수 있게 된다.At this time, the WHT is multiplied by 0 and 1, so that it can be implemented only by the addition calculation, which can significantly reduce the calculation amount and can reduce the correlation between adjacent samples of the input signal as an orthogonal transformation.
그리고 전력 정규화 과정을 통해 자기상관행렬의 고유치 퍼짐정도를 1에 가깝게 함으로써 신호의 특성이 변하는 구간에서도 신속한 적응성능이 나타나게 된다.In addition, through the power normalization process, the autocorrelation matrix's eigenvalue spreading value is close to 1, so that the adaptive performance is shown even in the region where the signal characteristic changes.
마지막으로 상기 단계(S6)에서 전력 정규화된 신호를 적응 예측기를 통해 적응 예측하여 출력신호 se(k)를 출력한다(S7).Finally, in step S6, the power normalized signal is adaptively predicted through the adaptive predictor to output the output signal s e (k) (S7).
이상, 상기 설명에서와 같이 본 발명은 WHT를 이용함에 따라 계산량을 줄일 수 있음은 물론 직교변환을 통해 입력신호의 인접 샘플간의 상관도를 줄일 수 있어 ADPCM을 위한 적응 예측기의 수렴 성능을 향상시킬 수 있게 되는 효과가 있다.As described above, the present invention can reduce the amount of calculation according to the use of the WHT, as well as the correlation between adjacent samples of the input signal through orthogonal transformation, thereby improving the convergence performance of the adaptive predictor for ADPCM. It is effective to be.
Claims (1)
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
KR1019970074054A KR19990054249A (en) | 1997-12-26 | 1997-12-26 | ADP-CM method using Walsh-Hadamard transformation |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
KR1019970074054A KR19990054249A (en) | 1997-12-26 | 1997-12-26 | ADP-CM method using Walsh-Hadamard transformation |
Publications (1)
Publication Number | Publication Date |
---|---|
KR19990054249A true KR19990054249A (en) | 1999-07-15 |
Family
ID=66165820
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
KR1019970074054A KR19990054249A (en) | 1997-12-26 | 1997-12-26 | ADP-CM method using Walsh-Hadamard transformation |
Country Status (1)
Country | Link |
---|---|
KR (1) | KR19990054249A (en) |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
KR100452509B1 (en) * | 2000-12-23 | 2004-10-12 | 엘지전자 주식회사 | Normalization method of signal power for a telecommunication system |
-
1997
- 1997-12-26 KR KR1019970074054A patent/KR19990054249A/en not_active Application Discontinuation
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
KR100452509B1 (en) * | 2000-12-23 | 2004-10-12 | 엘지전자 주식회사 | Normalization method of signal power for a telecommunication system |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
US5487086A (en) | Transform vector quantization for adaptive predictive coding | |
US8468017B2 (en) | Multi-stage quantization method and device | |
JPH11143499A (en) | Improved method for switching type predictive quantization | |
KR960006301A (en) | Sound signal encoding / decoding method | |
Shoham et al. | Efficient codebook allocation for an arbitrary set of vector quantizers | |
CN117194866B (en) | Distributed filtering method based on mass spring damping system | |
KR19990054249A (en) | ADP-CM method using Walsh-Hadamard transformation | |
KR100984637B1 (en) | Method and unit for substracting quantization noise from a pcm signal | |
Mumolo et al. | Adaptive predictive coding of speech by means of Volterra predictors | |
Chaddha et al. | Constrained and recursive hierarchical table-lookup vector quantization | |
JP2005516442A6 (en) | Method and unit for removing quantization noise from a PCM signal | |
Phamdo et al. | Coding of speech LSP parameters using TSVQ with interblock noiseless coding | |
Chen | Low-bit-rate predictive coding of speech waveforms based on vector quantization | |
KR100321705B1 (en) | Neural network nonlinear function converting apparatus and its method using power function of two | |
KR0138868B1 (en) | Lsp frequency quantizer | |
Liu et al. | Power spectra of ADPCM | |
Gibson | On the high rate, independence, and optimal prediction assumptions in predictive coding | |
Omara et al. | A compression-based backward approach for the forward sparse modeling with application to speech coding | |
Ramprasad et al. | Coding for low-power address and data busses: a source-coding framework and applications | |
Zhou et al. | Scientific data lossless compression using fast neural network | |
Tudela et al. | On the Convergence of the Structural Estimation of Proximal Operator with Gaussian Processes (STEP-GP) Method with Adaptive Quantization for Communication-Efficient Distributed Optimization | |
KR19990054252A (en) | Vocoder with Voice Signal Prediction by Volterra Filter | |
Sayood et al. | Explicit additive noise models for uniform and nonuniform MMSE quantization | |
Masgrau et al. | A continuously adaptive vector predictive coder (AVPC) for speech encoding | |
CN118101489A (en) | State estimation method, system and storage medium of nonlinear networked system based on asynchronous coding and decoding mechanism |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
WITN | Withdrawal due to no request for examination |