Frequency domain postfiltering for multiband excited linearpredictive coding of speech

A frequency domain postfilter for multiband excited linear predictive (MBELP) coders is proposed. It is based on the principle of multiplying the signal band magnitudes by the postfilter magnitudes obtained by sampling the postfilter response at the pitch harmonics. This approach offers two major advantages: first, the signal energies before and after postfiltering can be equalised by simply rescaling the signal band magnitudes; and secondly, the phase information will be retained owing to only the magnitude spectra being modified. A fast method is given for sampling the postfilter magnitude spectrum characterised by line spectral pair (LSP) parameters     

New scheme for vector quantisation of speech signals

The design theory of a vector quantiser is valid only on some assumptions which are unfortunately not satisfied in most real-world cases. A decomposition procedure of speech signals is described which leads to a proposed method for designing the vector quantisers in a reasonable way.<>     

Enhancement of ADPCM speech coding with backward-adaptivealgorithms for postfiltering and noise feedback

It is shown that postfiltering circuits based on higher order LPC (linear predictive coding) models can provide very low distortion in terms of special tilt. Thus, they can provide better speech enhancement than circuits based on the backward-adaptive pole-zero predictor in ADPCM (adaptive digital pulse code modulation). Quantitative criteria for designing postfiltering circuits based on higher-order LPC models are discussed. These postfilters are particularly attractive for systems where high-order LPC analysis is an integral part of the coding algorithm. In a subjective test that used a computer-simulated version of these circuits, enhanced ADPCM obtained a mean opinion score of 3.6 at 16 kb/s     

Pitch predictive vector quantisation of speech waveform

A coding algorithm is presented which combines pitch prediction with low-dimensional vector quantisation to exploit both long- and short-term correlation in the speech waveform at rates of 16 and 9.6 kbit/s. Vector quantisation of the predictor enables the stability of the synthesis filter to be assured, and also allows the use of a minimum residual energy criterion. SNRs of 17-19 dB are achieved at 16 kbit/s and 13-15 dB at 9.6 kbit/s.<>     

Optimal block boundary pre/postfiltering for wavelet-based image and video compression.

This paper presents a pre/postfiltering framework to reduce the reconstruction errors near block boundaries in wavelet-based image and video compression. Two algorithms are developed to obtain the optimal filter, based on boundary filter bank and polyphase structure, respectively. A low-complexity structure is employed to approximate the optimal solution. Performances of the proposed method in the removal of JPEG 2000 tiling artifact and the jittering artifact of three-dimensional wavelet video coding are reported. Comparisons with other methods demonstrate the advantages of our pre/postfiltering framework.     

Vector quantisation for minimum residual energy in pitch prediction of speech

A method for vector quantisation of pitch predictor coefficients according to a minimum residual energy criterion is proposed and compared to vector quantisation using the traditional minimum squared error between coefficients. Squared error quantisation is found to be adequate for 1-tap prediction, but for 3-tap prediction the residual energy method performs consistently better. The predictor code-books are also found to give robust performance outside the training sequence.     

Novel interpolative vector quantisation technique for image coding

An innovative technique for the interpolative vector quantisation (IVQ) of an image is developed. The technique employs a new split-and-merge algorithm which selects a near-to-optimal set of sampling points for the IVQ. Experimental results show that, under the same bit rate, over 0.5 dB improvement is achieved by the proposed method when compared with the results obtained by the traditional fixed sampling point interpolation in terms of peak signal-to-noise ratio (PSNR).<>     

Vector quantisation of wideband transform coefficient timetrajectories in speech coding

A number of methods to perform efficient dynamic bit allocation to perceptually important parameters in the frequency domain results from the combination of vector quantisation techniques and wideband transform coding. Time trajectories of transform coefficients and of magnitude and phase-derivative coefficients are used as vectors for vector quantisation. Speech coders implementable in single DSPs with external memory, having short delay and high robustness to bit stream distortions have been developed     

Digital image watermarking technique based on vector quantisation

A digital image watermarking technique based on vector quantisation (VQ) is presented. This technique uses codeword indices to carry the watermark information. The technique is secret and efficient: the watermarked image is robust to VQ compression with the same codebook. The simulation results prove the effectiveness of this technique     

Switched predictor adaptation of DPCM-AQB speech coder using an imbedded vector quantisation

An adaptive differential pulse-code-modulation speech coder with a switched predictor adaptation scheme is described. The predictor adaptation is done by switching one of the several predetermined predictors stored both at the encoder and the decoder. The selection of the predictor is controlled by a vector quantiser. The tradeoff between the performance and the extra side information that needs to be transmitted is evaluated for bit rates of 16, 24 and 32 Kbit/s.     

Low-complexity GMM-based block quantisation of images using the discrete cosine transform

While block transform image coding has not been very popular lately in the presence of current state-of-the-art wavelet-based coders, the Gaussian mixture model (GMM)-based block quantiser, without the use of entropy coding, is still very competitive in the class of fixed rate transform coders. In this paper, a GMM-based block quantiser of low computational complexity is presented which is based on the discrete cosine transform (DCT). It is observed that the assumption of Gaussian mixture components in a GMM having Gauss–Markov properties is a reasonable one with the DCT approaching the optimality of the Karhunen–Loève transform (KLT) as a decorrelator. Performance gains of 6–7 dB are reported over the traditional single Gaussian block quantiser at 1 bit per pixel. The DCT possesses two advantages over the KLT: being fixed and source independent, which means it only needs to be applied once; and the availability of fast and efficient implementations. These advantages, together with bitrate scalability, result in a block quantiser that is considerably faster and less complex while the novelty of using a GMM to model the source probability density function is still preserved.     

Parameter re-estimation in semicontinuous hidden Markov modellingof speech with feedback to vector quantisation codebook

The parameters of the semicontinuous hidden Markov model (SCHMM) can be re-estimated by allowing the codebook to be updated, thus achieving an optimised codebook/model combination. With the optimised codebook, the SCHMM can offer improved recognition accuracy in comparison to both the continuous and the discrete hidden Markov model     

Minimum-distortion functional for one-dimensional quantisation

The minimum rth-mean distortion resulting from quantisation of a one-dimensional signal is considered. In particular, the worst-case value of this minimum distortion is examined for signals satisfying rth-moment constraints. Numerical results are presented for the particular cases of minimum-mean-squared-error (r = 2) and minimum-mean-absolute-error (r = 1) quantisation.     

Colour quantisation technique based on image decomposition and its embedded system implementation

A new colour quantisation (CQ) technique and its corresponding embedded system realisation are introduced. The CQ technique is based on image split into sub-images and the use of Kohonen self-organised neural network classifiers (SONNC). Initially, the dominant colours of each sub-image are extracted through SONNCs and then are used for the quantisation of the colours of the entire image. The proposed CQ technique can use both colour components and spatial features, achieving better approximation of the final image to the spatial characteristics of the original one. In addition, for the estimation of the proper number of dominant image colours, a new algorithm based on the projection of the image colours into the first two principal components is proposed. The image split into sub-images offers reduction of the on-chip memory requirements and is suitable for embedded system (or system-on-chip) implementation of the most time-consuming part of the technique. Applying a systematic design methodology to the developed CQ algorithm, an efficient embedded architecture based on the ARM7 processor achieving high-speed processing and less energy consumption, is derived.     

Clustering-based quantisation for PDE-based image compression

Optimal known pixel data for inpainting in compression codecs based on partial differential equations is real-valued and thereby expensive to store. Thus, quantisation is required for efficient encoding. In this paper, we interpret the quantisation step as a clustering problem. Due to the global impact of each known pixel and correlations between spatial and tonal data, we investigate the central question, which kind of feature vectors should be used for clustering with popular strategies such as k-means. Our findings show that the number of colours can be reduced significantly without impacting the reconstruction quality. Surprisingly, these benefits are negated by an increased coding cost in compression applications.     

Fast clustering algorithm for vector quantisation

A fast clustering algorithm is presented as an alternative to the K-means algorithm. By encoding training vectors selectively and changing the codebook updating step, the algorithm reduces the computation time. Simulations show that the algorithm outperforms the K-means algorithm in computation time and performance in terms of mean-squared-error     

改进的高效动态时间规整算法语音识别系统

动态时间规整算法是结合了动态时间规整(DTW)技术和距离测度计算技术的一种非线性规整算法,在语音识别模板匹配中有重要的应用。为此提出一种改进的高效动态时间规整算法,其能有效加快搜索路径的寻找。基于Matlab实现了隐马尔科夫算法、高效动态时间规整算法和改进的高效动态时间规整算法的语音识别系统,同时进行了算法的仿真实验。实验结果表明,基于改进高效动态时间规整算法的训练速度远大于基于隐马尔可夫算法和高效动态时间规整算法的训练速度,而识别率下降很小,对于小词汇量非连续语音识别中高效动态时间规整算法的识别率为97.56%,隐马尔可夫算法的识别率为97.14%,改进高效动态时间规整算法的识别率为96.43%。     

Improved perceptually non-uniform spectral compression for robust speech recognition

According to the decline of recognition rate of speech recognition system in the noise environments, an improved perceptually non-uniform spectral compression feature extraction algorithm is put forward in this paper. This method can realize an effective compression of the speech signals and make the training and recognition environments more matching, so the recognition rate can be improved in the noise environments. By experimenting on the intelligent wheelchair platform, the result shows that the algorithm can effectively enhance the robustness of speech recognition, and ensure the recognition rate in the noise environments.     

Efficient algorithm for channel-optimised vector quantisation

An efficient method is presented for increasing the speed of the clustering process for channel-optimised vector quantisation (COVQ), in which the Hamming distances between the indices of codevectors are utilised. Simulation results demonstrate that with a small increase in pre-processing and memory costs, the training time of the new algorithm has been reduced significantly while the encoding quality remains unchanged     

Fast search algorithm for vector quantisation

A tree structure for the fast nearest neighbour search algorithm for vector quantisation is presented. Using the new algorithm a remarkable reduction in the number of multiplications, additions and comparisons is achieved.<>