This paper describes extensions to trellis-based scalar-vector quantization (TB-SVQ) , as novel techniques for coding sources with memory. These techniques are proposed as efficient solutions for the quantization of the excitation in Code-Excited Linear Predictive (CELP) coding algorithms. A new 24 kbit/s low-delay trellis CELP (LD-TCELP) coder is introduced which achieves high-performance coding of narrowband speech with an algorithmic delay of only 3 ms. This coder utilizes the granular structure of the underlying trellis code and the codebook shaping capability of the SVQ approach to effectively capitalize on the gain of higher dimensional (24-D) vector quantization. We present experimental results which show that the LD-TCELP coder outperforms two toll-quality standardized coders: ITU-T G.728 16 kbit/s LD-CELP and ITU-T G.726 32 kbit/s ADPCM. The proposed LD-TCELP coder can also utilize a low-complexity codebook search approach with a computational complexity that is about 75% of that of ITU-T G.728 LD-CELP.

Many modern analog media coders employ some form of entropy coding (EC). Usually, a simple per-letter EC is used to keep the coder's complexity and price low. In some coders, individual symbols are grouped into small fixed-size vectors before EC is applied. We extend this approach to form variable-size vector EC (VSVEC) in which vector sizes may be from 1 to several hundreds. The method is, however, complexity-constrained in the sense that the vector size is always as large as allowed by a pre-set complexity limit. The idea is studied in the framework of a modified discrete cosine transform (MDCT) coder. It is shown experimentally, using diverse audio material, that a rate reduction of about 37% can be achieved. The method is, however, not specific to MDCT coding but can be incorporated in various speech, audio, image and video coders.

This paper describes the 8 kb/s speech coding algorithm G.729 which has been standardized by ITU-T. The algorithm is based on a conjugate-structure algebraic CELP (CS-ACELP) coding technique and uses 10 ms speech frames. The codec delivers toll-quality speech (equivalent to 32 kb/s ADPCM) for most operating conditions. This paper describes the coder structure in detail and discusses the reasons behind certain design choices. A 16-b fixed-point version has been developed as part of Recommendation G.729 and a summary of the subjective test results based on a real-time implementation of this version are presented.

The recently-introduced waveform interpolation (WI) coders provide good-quality speech at low rates but may be too complex for commercial use. This paper proposes new approaches to low-complexity WI speech coding at rates of 1.2 and 2.4 kbps. The proposed coders are 4 to 5 times faster than the previously reported ones. At 2.4 kbps, the complexity is about 7.5 and 2.5 MFLOPS for the encoder and decoder, respectively. At 1.2 kbps, the complexity is about 6 and 2.3 MFLOPS for the encoder and decoder, respectively. Informal subjective evaluation shows that, at 2.4 kbps, the quality is close to that of the high-complexity coders. The quality does not significantly degrade at 1.2 kbps and it is considered sufficient for messaging applications.

A recent independent survey found a 2.4 kbit/s waveform-interpolation (WI) algorithm to perform better than other state-of-the-art speech coders. However, this coder had a very high level of computational complexity. The introduction of various techniques, including a time-varying waveform sampling rate and a cubic B-spline waveform representation, has reduced the computational complexity by an order of magnitude. The new implementation allows full-duplex real-time operation on a single DSP device and on an average workstation (the latter using non-optimized compiled C source code). The new coder also contains a number of new features which improve the quality of the reconstructed speech signal.

In this paper we propose a simple fast-search VQ of the LSF's to be used on top of the split VQ, that is, in each of the sub-vector domains.The main trait of the proposed method is that no suboptimal codebooks are used and there is no further reduction in performance.In each sub-vector domain, a full-size optimally trained codebook, typically of size 1024, is searched using a fastsearch algorithm.The result of this search is identical to that of a full-search, yet, only about 25% of a full-search complexity is needed.

The Wideband-Audio Low-Delay CELP (LD-CELP) coder produces speech with quality as high as the CCITT 64 kb/s standard (G.722) at half the bitrate. The computational load of the encoder is almost 900% processor time of the 12.5 MIPS DSP32c. This makes a real-time implementation impractical. We investigated the Gain-Shape Vector-Quantization (GSVQ) in order to reduce the computational load of the encoder. This paper describes a real-time implementation of the LD-CELP encoder based on the AT&T SURFboard using two DSP32c operating in parallel. A computational load of 180% processor time has been achieved. The respective decoder requires 42% processor time. The implementation of a full-duplexed coding system requires three 12.5 MIPS Digital-Signal-Processors (DSPs) and has oneway coding delay of less than 1ms. The coder also performs well for non-speech wideband audio signals such as music.

The authors report on the use of the codebook-excited linear-predictive (CELP) algorithm for 32 kb/s low-delay (LD-CELP) coding of wideband speech. The main problem associated with wideband coding, namely, spectral noise weighting, is discussed. The authors propose an enhanced noise weighting technique and demonstrate its efficiency via subjective listening tests. In these tests, involving 20 listeners and 8 test sentences, the average rating for the proposed 32 kb/s LD-CELP was essentially equal to that of the 65 kb/s standard (G.722) CCITT wideband coder.>

It is shown that, within the framework of LPC (linear predictive coding)-based analysis-by-synthesis coding, the codebook search can be implemented without the filtering operation, thus alleviating the complexity problem and allowing the quantizer to be structured for fast-search quantization. In the proposed scheme, the vector quantizer bypasses the filter and direct-vector-quantization (DVQ) is used for quantizing the input vectors. It is demonstrated that this algorithm performs as well as (if not better than) the conventional analysis-by-synthesis system for short-frame (low delay) applications. Although the DVQ may be significantly less complex (depending on the type of fast-search VQ used), its signal-to-noise ratio and perceptual performance are the same as those of the conventional low-delay CELP (codebook-excited linear prediction).>

Cascade likelihood vector quantization is proposed for coding LPC spectral parameters for application in low-rate code-excited linear prediction systems. The approach is based on representing the LPC all-pole filter as a cascade of two lower-order all-pole filters. The partitioning of the LPC polynomial is done in the root domain by clustering the roots into two distinct groups. The quantizer uses two codebooks to quantize each of the lower-order filters. However, the quantization process is done so as to optimize jointly the performances of the two subsystems. The likelihood-ratio distortion measure is used as a performance criterion in the design and coding processes. Splitting the LPC filter into two subsystems dramatically reduces the coding complexity while the efficiency of using vector quantization is essentially preserved. Experimental results show an average performance of 1.59, 1.41, 1.24, and 1.10 dB log-spectral distortion at the rates of 1.0, 1.1, 1.2, and 1.3 kb/s, respectively. These figures represent good quality at 1.0 kb/s up to essentially transparent quality at 1.3 kb/s.>

Two very different subband coders are described. The first is a modified dynamic bit-allocation-subband coder (D-SBC) designed for variable rate coding situations and easily adaptable to noisy channel environments. It can operate at rates as low as 12 kb/s and still give good quality speech. The second coder is a 16-kb/s waveform coder, based on a combination of subband coding and vector quantization (VQ-SBC). The key feature of this coder is its short coding delay, which makes it suitable for real-time communication networks. The speech quality of both coders has been enhanced by adaptive postfiltering. The coders have been implemented on a single AT&T DSP32 signal processor.>

Vector Predictive Quantization (VPQ) is proposed for coding the short-term spectral envelope of speech. The proposed VPQ scheme predicts the current spectral envelope from several past spectra, using a predictor codebook. The residual spectrum is coded by a residual codebook. The system operates in the log-spectral domain using a sampled version of the spectral envelope. Experimental results indicate a prediction gain in the range of 9 to 13 dB and an average log-spectral distance of 1.3 to 1.7 dB. Informal listening tests suggest that replacing the conventional scalar quantizer in a 4.8 Kbits/s CELP coder by a VPQ system allows a reduction of the rate assigned to the LPC data from 1.8 Kbits/s to 1.0 Kbits/s without any obvious difference in the perceptual quality.

Traditional solutions to the bit allocation problem assume nicely behaved quantizers whose distortion versus rate characteristic is a standard exponentially decreasing function. This model is often inadequate and leads to suboptimal bit allocations for actual quantizer characteristics. This inadequacy prevails particularly in coding systems based on vector quantization (VQ). The distortion-rate characteristic for VQ suffers from irregularities because of suboptimality of the codebook and of the codebook search method. We propose a new algorithm which efficiently allocates a given quota of bits, based on the actually measured quantizer performance, without any prior assumptions on its behavior. The algorithm can enhance the performance-of traditional scalar coders as well as any modern coder which employs dynamic bit allocation.

A fundamental computational task that arises in several areas of signal processing is pattern matching, where a given test pattern is compared with a large set of stored templates, to find the best match that minimizes a given measure of dissimilarity. Three different geometrically-oriented methods are proposed for substantially reducing the computational complexity of the search process by reducing the number of multiplies in exchange for additional low complexity operations and, in two of the methods, additional memory for storing precomputed tables.

This paper introduces a Hierarchical Vector Quantization (HVQ) scheme that can operate on "supervectors" of dimensionality in the hundreds of samples. HVQ is based on a tree-structured decomposition of the original super-vector into a large number of low dimensional vectors. The supervector is partitioned into subvectors, the subvectors into minivectors and so on. The "glue" that links subvectors at one level to the parent vector at the next higher level is a feature vector that characterizes the correlation pattern of the parent vector and controls the quantization of lower level feature vectors and ultimately of the final descendant data vectors. Each component of a feature vector is a scalar parameter that partially describes a corresponding subvector. The paper presents a three level HVQ for which the feature vectors are based on subvector energies. Gain normalization and dynamic codebook allocation are used in coding both feature vectors and the final data subvectors. Simulation results demonstrate the effectiveness of HVQ for speech waveform coding at 9.6 and 16 Kb/s.