A finite-state entropy-constrained vector quantizer for audio MDCT coefficients coding

Abstract

In this paper, an entropy-constrained vector quantizer (ECVQ) scheme with finite memory called finitestate ECVQ (FS-ECVQ) is presented for saving the large memory requirements and improving the coding performance of an ordinary vector quantizer (VQ). This quantizer consists of a finite-state vector quantizer (FSVQ) and multiple ECVQs. The source sequence is first split into multiple clusters by the FSVQ. Then, to each cluster a dedicated ECVQ is applied. By the FSVQ, the FS-ECVQ effectively exploits the inter-frame dependencies, and moreover, there is no need to transmit the cluster indexes to the decoder side. By ECVQs, the ratedistortion (R/D) performance and the total memory requirements of the FS-ECVQ are able to be efficiently improved and decreased, respectively. A FS-ECVQ was implemented and evaluated based on USAC work draft 6 coding scheme. Results showed that compared to the original encoder, the proposed quantizer cut down the total memory requirements of 87% and improved the average coding gain to nine items of 3.87%, simultaneously.

Finite‐state entropy‐constrained vector quantiser for audio modified discrete cosine transform coefficients uniform quantisation

“…Compared with the statistical parameter used in [17], the new parameter provides a more simpler and more accurate description of the fact that the shape feature is more closely related to the block energy than to the block skewness.…”

“…In this paper, a composite quantiser, called FS‐ECVQ, is introduced, in which multiple ECVQs are combined with a FSVQ. In FS‐ECVQ [17], the FSVQ serves as a classifier which splits the source sequence into multiple clusters. To achieve better classification performance, the FSVQ employs multiple previous adjacent vectors, even including those within previous frames, to draw current decision, and therefore better exploits the inter‐frame redundancies than an ordinary SVQ does.…”

“…Thus, the FS‐ECVQ will be more robust than an ordinary SVQ. In addition, compared with [17], our new method employs a new statistical parameter, which describes the shape feature of an input block in a more simpler way, and adopts an iteration method to merge the similar pdf models, which obviously decreased the total memory requirements.…”

A memory efficient finite-state source coding algorithm for audio MDCT coefficients

“…In this algorithm, a block, x, is first separated from the source sequence, whose length is set to be the largest available vector dimension, supposed to be 8. Then, the current state s of the obtained block x, which is calculated through (23), is compared with a given threshold, T 8 . If current state s is lower than T 8 , block x will be taken as an element belonging to cluster 8 .…”

“…In FS-ECVQ [23], this FSVQ serves as a classifier which splits the source sequence into multiple clusters. To achieve better classification performance, the FSVQ draws the current decision based on information obtained from a number of previous adjacent vectors, even from those in previous frames, and thus better exploits the inter-frame redundancies than an ordinary SVQ does.…”