Linearization of the Optimal Compressor Function for Gaussian Mixture Model Zoran H. Perić, Jelena R. Nikolić, Vladica N. Đor đ ević.

Slides:



Advertisements
Similar presentations
A COMPARATIVE STUDY OF DCT AND WAVELET-BASED IMAGE CODING & RECONSTRUCTION Mr. S Majumder & Dr. Md. A Hussain Department of Electronics & Communication.
Advertisements

OFDM Transmission over Gaussian Channel
Pulse Code Modulation Pulse Code Modulation
Feature Selection as Relevant Information Encoding Naftali Tishby School of Computer Science and Engineering The Hebrew University, Jerusalem, Israel NIPS.
Sampling and Pulse Code Modulation
Digital Coding of Analog Signal Prepared By: Amit Degada Teaching Assistant Electronics Engineering Department, Sardar Vallabhbhai National Institute of.
Lecture 25 Pulse-Width Modulation (PWM) Techniques
Digital Image Processing
Quantization Prof. Siripong Potisuk.
Pattern Classification All materials in these slides were taken from Pattern Classification (2nd ed) by R. O. Duda, P. E. Hart and D. G. Stork, John Wiley.
Chinese University of Hong Kong Department of Information Engineering A Capacity Estimate Technique for JPEG-to-JPEG Image Watermarking Peter Hon Wah Wong.
Digital Voice Communication Link EE 413 – TEAM 2 April 21 st, 2005.
1 A Unified Rate-Distortion Analysis Framework for Transform Coding Student : Ho-Chang Wu Student : Ho-Chang Wu Advisor : Prof. David W. Lin Advisor :
Matched Filters By: Andy Wang.
Jacinto C. Nascimento, Member, IEEE, and Jorge S. Marques
Waveform SpeechCoding Algorithms: An Overview
296.3Page :Algorithms in the Real World Convolutional Coding & Viterbi Decoding.
Fundamentals of Digital Communication
Image coding/compression
IMAGE SAMPLING AND IMAGE QUANTIZATION 1. Introduction
Rate-distortion modeling of scalable video coders 指導教授:許子衡 教授 學生:王志嘉.
ECE 4710: Lecture #9 1 PCM Noise  Decoded PCM signal at Rx output is analog signal corrupted by “noise”  Many sources of noise:  Quantizing noise »Four.
Digital Signal Processing
COMMON EVALUATION FINAL PROJECT Vira Oleksyuk ECE 8110: Introduction to machine Learning and Pattern Recognition.
International Conference on Intelligent and Advanced Systems 2007 Chee-Ming Ting Sh-Hussain Salleh Tian-Swee Tan A. K. Ariff. Jain-De,Lee.
2010/12/11 Frequency Domain Blind Source Separation Based Noise Suppression to Hearing Aids (Part 1) Presenter: Cian-Bei Hong Advisor: Dr. Yeou-Jiunn Chen.
1 Security and Robustness Enhancement for Image Data Hiding Authors: Ning Liu, Palak Amin, and K. P. Subbalakshmi, Senior Member, IEEE IEEE TRANSACTIONS.
CMPT 365 Multimedia Systems
CE Digital Signal Processing Fall 1992 Waveform Coding Hossein Sameti Department of Computer Engineering Sharif University of Technology.
MPEG Audio coders. Motion Pictures Expert Group(MPEG) The coders associated with audio compression part of MPEG standard are called MPEG audio compressor.
EEE 503 Digital Signal Processing Lecture #1 : Introduction Dr. Panuthat Boonpramuk Department of Control System & Instrumentation Engineering KMUTT.
1 PCM & DPCM & DM. 2 Pulse-Code Modulation (PCM) : In PCM each sample of the signal is quantized to one of the amplitude levels, where B is the number.
Digital image processing Chapter 3. Image sampling and quantization IMAGE SAMPLING AND IMAGE QUANTIZATION 1. Introduction 2. Sampling in the two-dimensional.
Tamal Bose, Digital Signal and Image Processing © 2004 by John Wiley & Sons, Inc. All rights reserved. Figure 8-1 (p. 491) Adaptive channel equalizer.
Design of Novel Two-Level Quantizer with Extended Huffman Coding for Laplacian Source Lazar Velimirović, Miomir Stanković, Zoran Perić, Jelena Nikolić,
Smooth Side-Match Classified Vector Quantizer with Variable Block Size IEEE Transaction on image processing, VOL. 10, NO. 5, MAY 2001 Department of Applied.
1 Quantization Error Analysis Author: Anil Pothireddy 12/10/ /10/2002.
Outline Transmitters (Chapters 3 and 4, Source Coding and Modulation) (week 1 and 2) Receivers (Chapter 5) (week 3 and 4) Received Signal Synchronization.
CHAPTER 5 SIGNAL SPACE ANALYSIS
Introduction to Digital Signals
PCM & DPCM & DM.
Computer simulation Sep. 9, QUIZ 2 Determine whether the following experiments have discrete or continuous out comes A fair die is tossed and the.
The Fractional Fourier Transform and Its Applications Presenter: Pao-Yen Lin Research Advisor: Jian-Jiun Ding, Ph. D. Assistant professor Digital Image.
Image Coding/ Compression
ELE 488 F06 ELE 488 Fall 2006 Image Processing and Transmission ( ) Image Compression Review of Basics Huffman coding run length coding Quantization.
Chapter 8 Lossy Compression Algorithms. Fundamentals of Multimedia, Chapter Introduction Lossless compression algorithms do not deliver compression.
Performance of Digital Communications System
1 Introduction to Speech Coding What, Why, Where & How (First Part) By Allam Mousa Department of Telecommunication Engineering An Najah University SP_2_Coding_1of2.
DIGITAL COMMUNICATION. Introduction In a data communication system, the output of the data source is transmitted from one point to another. The rate of.
Efficient Huffman Decoding Aggarwal, M. and Narayan, A., International Conference on Image Processing, vol. 1, pp. 936 – 939, 2000 Presenter :Yu-Cheng.
EKT 431 DIGITAL COMMUNICATIONS. MEETING LECTURE : 3 HOURS LABORATORY : 2 HOURS LECTURER PUAN NORSUHAIDA AHMAD /
بسم الله الرحمن الرحيم Digital Signal Processing Lecture 2 Analog to Digital Conversion University of Khartoum Department of Electrical and Electronic.
VIDYA PRATISHTHAN’S COLLEGE OF ENGINEERING, BARAMATI.
Chapter 8 Lossy Compression Algorithms
Analog to digital conversion
Outlier Processing via L1-Principal Subspaces
UNIT II.
Sampling rate conversion by a rational factor
Context-based Data Compression
Scalar Quantization – Mathematical Model
PCA vs ICA vs LDA.
Soutenance de thèse vendredi 24 novembre 2006, Lorient
MOTION ESTIMATION AND VIDEO COMPRESSION
PCM & DPCM & DM.
The Fractional Fourier Transform and Its Applications
ELEC6111: Detection and Estimation Theory Course Objective
Conceptual Representation of A/D Conversion
Fixed-point Analysis of Digital Filters
INTRODUCTION TO DIGITAL COMMUNICATION
Combination of Feature and Channel Compensation (1/2)
Presentation transcript:

Linearization of the Optimal Compressor Function for Gaussian Mixture Model Zoran H. Perić, Jelena R. Nikolić, Vladica N. Đor đ ević

In this paper, the linearization of the optimal compressor functions were done by using a method with a variable number of representational levels within segments. Design of linearized compandor for the proposed probability density function (GMM) was done, followed by analysis of the results. The paper is structured as following: 1. Introduction 2. GMM (Gaussian mixture model) 3. Linearization of the Optimal Compressor Function 4. Numerical Results 5. Conclusion

1. Introduction Scalar quantization. Uniform and non-uniform quantizers. Compression as the equivalent to a non-uniform quantization (Bennet). Which is the most commonly used compressor functions? What is the advantage of optimal compressor function?

2. GMM (Gaussian mixture model) In order to achieve a better approximation of the histogram of probability density function of real signals, than in the case of Laplacian and Gaussian sources, the Gaussian mixture model, was used. In this case, the Gaussian mixture model, which consists of two Gaussian components, was used:

Fig. 1. GMM in the case of different mean values of Gaussian components

3. Linearization of the Optimal Compressor Function Because of the very complex practical realization of compressor function, its linearization must be performed. Optimal compressor function: Optimal compressor function for the proposed Gaussian mixture model:

Fig. 2. Illustration of the linearized optimal compressor function

Segments are determined using the following expression: where L-is the number of segments. Number of cells within the segments: where c i and c i-1 - are the output values of the compressor for the corresponding values t i and t i-1, respectively.

Size of the amplitude quantum: Granular distortion in the case of linearized models: Probability that the current value of the input signal belongs to the i-th segment:

Overload distortion: where y ov - is the representational level with the highest value, Total distortion:

SQNR (Signal-to-quantization-noise ratio): where σ²-is the variance of the input signal,

4. Numerical Results Fig. 3. Optimal compressor functions for three different mean values ​​ of the Gaussian component, m 2.

Piecewise linear compressor fuction for m2=1

Piecewise linear compressor fuction for m2=4

Piecewise linear compressor fuction for m2=8

Table I Values of SQNR for different number of representational levels of quantizer and for different mean values of Gaussian components 5. Conclusion The significance of this work lies in the fact that in order to achieve a better approximation of the histogram of probability density function of real signals, than in the case of Laplacian and Gaussian sources [7,8], the Gaussian mixture model, which consists of two Gaussian components, was used [4]. This is reflected in the significantly higher value of SQNR than in the case of linearization of the optimal compressor functions reported in [7,8].

References [1] A. Gersho and R. M. Gray, Vector Quantization and Signal Compression. Boston, Dordrecht, London: Kluwer Academic Publishers, 1992, ch. 5, pp [2] J. D. Gibson, The Communications Handbook. New York: Crc Press Inc, 2002, 2nd ed., ch. 3, pp [3] L. R. Rabiner and R. W. Schafer, "Introduction to Digital Speech Processing", Foundations and Trends in Signal Processing, vol. 1, pp , Jan [4] N. S. Jayant and P. Noll, Digital Coding Of Waveforms, Principles and Applications to Speech and Video. New Jersey: Prentice Hall, 1984, 2nd ed., chs. 4-5, pp [5] W. C. Chu, Speech Coding Algorithms, Foundation and Evolution of Standardized Coders. New Jersey: John Wiley & Sons, 2003, ch 5-6, pp [6] Y. Q. Shi and H. Sun, Image and Video Compression for Multimedia Engineering: Fundamentals, Algorithms, and Standards. New York: Crc Press Inc, 2008, 2nd ed., ch. 2, pp [7] Z. Perić, J. Nikolić, "Linearizacija optimalne kompresorske funkcije", Zbornik radova XVII telekomunikacionog foruma TELFOR 2009, CD izdanje, str , Beograd, novembar, [8] Z. Perić, J. Nikolić, Z. Eskić, "Projektovanje linearizovanog hibridnog skalarnog kvantizera za Gausov izvor", Zbornik radova INFOTEH JAHORINA 2009, Vol.8, Ref. B- I-10, str , Jahorina, mart, 2009.

Thank you for your attention!