This paper describes an ITU-T G.711 embedded wideband speech coder, submitted as a candidate to the ITU-T G.711 wideband extension standardization qualification phase. The codec generates a bitstream comprised of three layers: a G.711 core layer with noise shaping, a time-domain weighted vector quantized narrowband enhancement layer, and an MDCT-based higher band enhancement layer. Through subjective evaluations, the coder was found to meet all tested requirements and objectives set in terms of reference, with a low computational complexity at 10 WMOPS.

A robust dereverberation method is presented for speech enhancement in a situation requiring adaptation where a speaker shifts his/her head under reverberant conditions causing the impulse responses to change frequently. We combine correlation-based blind deconvolution with modified spectral subtraction to improve the quality of inverse-filtered speech degraded by the estimation error of inverse filters obtained in practice. Our method computes inverse filters by using the correlation matrix between input signals that can be observed without measuring room impulse responses. Inverse filtering reduces early reflection, which has most of the power of the reverberation, and then, spectral subtraction suppresses the tail of the inverse-filtered reverberation. The performance of our method in adaptation is demonstrated by experiments using measured room impulse responses. The subjective results indicated that this method provides superior speech quality to each of the individual methods: blind deconvolution and spectral subtraction.

A dereverberation technique is developed to provide an alternative means of reducing reverberation in speech signals. The conventional MINT (the multiple-input/output inverse-filtering theorem) method uses the room impulse responses to calculate the inverse filters, so it cannot recover speech signals in practice, where the room impulse responses are unknown in advance. Our method blindly estimates the inverse filters by computing the correlation matrix between input signals that can be observed, instead of room impulse responses. We also combine the inverse filtering with modified spectral subtraction against the estimation error of inverse filters used in the field. The performance of the proposed method is demonstrated using actual room impulse responses

We propose a novel step-size control method that treats the echo path imbalance problem in stereo echo cancellation. The imbalance between two echo paths in magnitude slows down the convergence rate of the adaptive filter for stereo echo cancellation. To speed up the convergence rate, the proposed method extracts the information about filter-coefficient errors of each unknown echo path, and controls the step-size of each channel accordingly.

This paper presents a way to calculate the perceived importance of speech segments as a single value criterion, using a linear regression model. Unlike the commonly used voice activity detection (VAD) algorithms, this method allows us to obtain a finer priority granularity of speech segments. This can be used in conjunction with frequency scalable speech coding techniques and IP QoS techniques to achieve efficient and qualitycontrolled voice transmission. A simple linear regression model is used to calculate the estimated mean opinion score (MOS) of the various cases of missing speech segments.

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.

A wideband speech scalable codec is proposed for improving the flexibility in telecommunication networks. This coder is scalable with G.729 (ITU 8-kbit/s standard). Its decoder can process the incoming bitstream at three bit rates (8, 12, and 16 kbit/s) and provide a choice of speech types (wideband and telephone-band). The codec has a split-band structure, where both bands are coded by analysis-by-synthesis techniques. This paper proposes two types of scalable codec: a separate one and a composite one. It also proposes a new method (an additional adaptive codebook) for predicting pitch, while maintaining scalability with the G.729 codec. Subjective testing for wideband speech showed that the quality of the proposed codec at 16-kbit/s is equivalent to that of the 64-kbit/s G.722, and at 12-kbit/s is better than that of the 48-kbit/s G.722. Testing has further demonstrated that the 8-kbit/s coder provides high quality for telephone-band speech.

Low-cost highly-efficient speech coding is important for personal multimedia communications. In this paper we propose a new low-complexity speech coding method called "Dual-Pulse CS-CELP (DP-CS-CELP)" at 7.8 kbit/s. This method is based on ITU-T G.729. To reduce the complexity, we applied a new excitation model to the random code vectors and simplified the LPC coding, adaptive codebook search, perceptual weighting, and the other structures. The number of operations for this encoder is 3.80 MOPS, which achieves real-time speech coding and decoding on personal computers. Although the efficiency of this coder is a little lower than that of G.729, the MOS listening test showed that the subjective quality was equivalent to or a little better than that of G.726 (32-kbit/s ADPCM).

This paper describes a high-quality 8-kb/s speech coder called conjugate structure code-excited linear prediction (CS-CELP) with a 10-ms frame length. To provide a short delay and high quality under both error-free and channel error conditions, it uses three new schemes: line spectrum pair (LSP) quantization using interframe prediction, preselection in the codebook search, and gain vector quantization (VQ) with backward prediction. The LSP parameters are quantized by using multistage VQ with moving-average (MA) prediction. This scheme can operate efficiently with various frequency responses of speech. The preselection of the codebook reduces the computational complexity and improves the robustness to channel errors. The gain VQ with backward prediction can provide a high quality and robustness without transmission of input speech power information. A conjugate structure for both random codebook and gain codebook is introduced to improve the ability to handle random bit errors and to reduce codebook storage memory requirements. Subjective testing indicates that the quality of this coder is equivalent to that of 32-kb/s adaptive differential pulse code modulation (ADPCM) under error-free conditions. Testing has further demonstrated that the coder is robust against random bit errors.

A high-quality 8-kbit/s speech coder based on conjugate structure CELP (CS-CELP) is proposed that uses a trained sparse conjugate codebook. The trained sparse conjugate codebook improves speech quality for noisy speech. This codebook consists of two sub-codebooks and each sub-codebook consists of a random component and a trained component. Each component has excitation vectors consisting of a few pulses. In the random component, pulse position and amplitude are determined randomly. The trained component is determined by training. Subjective tests (differential mean opinion score, DMOS and mean opinion score, MOS) indicated that this codebook improves speech quality compared with the conventional trained codebook for noisy speech. The MOS showed that the quality of improved CS-CELP is equivalent to that of the 32-kbit/s ADPCM for clean speech.

This paper presents a high-quality 8-kbit/s speech coder (conjugate structure CELP: CS-CELP) that is a candidate for standardization by the ITU-T (formerly CCITT). To achieve high-quality for two types of speech (IRS and non-IRS (flat) speech) and real-time implementation, CS-CELP has been revised by two novel schemes. To handle two types of speech, the LSP parameters are quantized by multistage VQ with fourth-order interframe MA prediction. This scheme has little spectrum distortion, even if the two types of speech have many variations of the LSP parameters. The computational complexity of the implementation is reduced for adaptive and fixed-shape codebooks without degrading the speech quality. Multistage selection is adopted in the adaptive codebook; this selection uses a truncated impulse response. Improved pre-selection is proposed in the fixed-shape codebook. Subjective testing indicates that the quality of CS-CELP is equivalent to that of the 32-kbit/s ADPCM under error-free conditions for IRS and non-IRS speech. It also operates in real time using fixed-point DSP chips.>