Exploiting spatial and temporal redundancies for vector quantization of speech and images

Vector quantization;Speech coding;Image coding;Source coding;Codebook reordering;Signal processing

Meh Chu, Chu ; Anderson, David V. Electrical and Computer Engineering Moore II, Elliot Morley, Thomas Yezzi, Anthony Romberg, Justin K. ; Anderson, David V.

University:Georgia Institute of Technology

Department:Electrical and Computer Engineering

关键词: Vector quantization;    Speech coding;    Image coding;    Source coding;    Codebook reordering;    Signal processing;   

Others  :  https://smartech.gatech.edu/bitstream/1853/54442/1/MEHCHU-DISSERTATION-2015.pdf

美国|英语

来源: SMARTech Repository

【 摘 要 】

The objective of the proposed research is to compress data such as speech, audio, andimages using a new re-ordering vector quantization approach that exploits the transitionprobability between consecutive code vectors in a signal. Vector quantization is the processof encoding blocks of samples from a data sequence by replacing every input vector froma dictionary of reproduction vectors. Shannon’s rate-distortion theory states that signalsencoded as blocks of samples have a better rate-distortion performance relative to whenencoded on a sample-to-sample basis. As such, vector quantization achieves a lower codingrate for a given distortion relative to scalar quantization for any given signal.Vector quantization does not take advantage of the inter-vector correlation between successiveinput vectors in data sequences. It has been demonstrated that real signals have significantinter-vector correlation. This correlation has led to vector quantization approachesthat encode input vectors based on previously encoded vectors.Some methods have been proposed in literature to exploit the dependence between successivecode vectors. Predictive vector quantization, dynamic codebook re-ordering, andfinite-state vector quantization are examples of vector quantization schemes that use intervectorcorrelation. Predictive vector quantization and finite-state vector quantization predictthe reproduction vector for a given input vector by using past input vectors. Dynamiccodebook re-ordering vector quantization has the same reproduction vectors as standardvector quantization. The dynamic codebook re-ordering algorithm is based on the conceptof re-ordering indices whereby existing reproduction vectors are assigned new channel indicesaccording a structure that orders the reproduction vectors in an order of increasingdissimilarity. Hence, an input vector encoded in the standard vector quantization methodis transmitted through a channel with new indices such that 0 is assigned to the closestreproduction vector to the past reproduction vector. Larger index values are assigned toreproduction vectors that have larger distances from the previous reproduction vector.Dynamic codebook re-ordering assumes that the reproduction vectors of two successivevectors of real signals are typically close to each other according to a distance metric.Sometimes, two successively encoded vectors may have relatively larger distances fromeach other. Our likelihood codebook re-ordering vector quantization algorithm exploitsthe structure within a signal by exploiting the non-uniformity in the reproduction vectortransition probability in a data sequence. Input vectors that have higher probability of transitionfrom prior reproduction vectors are assigned indices of smaller values. The codevectors that are more likely to follow a given vector are assigned indices closer to 0 whilethe less likely are given assigned indices of higher value. This re-ordering provides thereproduction dictionary a structure suitable for entropy coding such as Huffman and arithmeticcoding. Since such transitions are common in real signals, it is expected that ourproposed algorithm when combined with entropy coding algorithms such binary arithmeticand Huffman coding, will result in lower bit rates for the same distortion as a standardvector quantization algorithm.The re-ordering vector quantization approach on quantized indices can be useful inspeech, images, audio transmission. By applying our re-ordering approach to these datatypes, we expect to achieve lower coding rates for a given distortion or perceptual quality.This reduced coding rate makes our proposed algorithm useful for transmission and storageof larger image, speech streams for their respective communication channels. The useof truncation on the likelihood codebook re-ordering scheme results in much lower compressionrates without significantly distorting the perceptual quality of the signals. Today,texts and other multimedia signals may be benefit from this additional layer of likelihoodre-ordering compression.