Steffen Duus Hansen

Reduction of broad-band noise in speech by truncated QSVD

We consider an algorithm for reduction of broadband noise in speech based on signal subspaces. The algorithm is formulated by means of the quotient singular value decomposition (QSVD). With this formulation, a prewhitening operation becomes an integral part of the algorithm. We demonstrate that this is essential in connection with updating issues in real-time recursive applications. We also illustrate by examples that we are able to achieve a satisfactory quality of the reconstructed signal.

Functional neuromuscular stimulation controlled by surface electromyographic signals produced by volitional activation of the same muscle: adaptive removal of the muscle response from the recorded EMG-signal

In order to use the volitional electromyography (EMG) as a control signal for the stimulation of the same muscle, it is necessary to eliminate the stimulation artifacts and the muscle responses caused by the stimulation. The stimulation artifacts, caused by the electric field in skin and tissue generated by the stimulation current, are relatively easy to eliminate by shutting down the EMG-amplifier at the onset of the stimulation pulses. The muscle response is a nonstationary signal, therefore, an adaptive linear prediction filter is proposed. The filter is implemented and for three filter lengths tested on both simulated and real data. The filter performance is compared with a conventional fixed comb filter. The simulations indicate that the adaptive filter is relatively insensitive to variations in amplitude of the muscle responses, and for all filter lengths produces a good filtering. For variations in shape of the muscle responses and for real data, an increased filter performance can be achieved by increasing the filter length. Using a filter length of up to seven stimulation periods, it is possible to reduce real muscle responses to a level comparable with the background noise. Using the shut-down circuit and the adaptive filter both the stimulation artifacts and the muscle responses can be effectively eliminated from the EMG signal from a stimulated muscle. It is therefore possible to extract the volitional EMG from a partly paralyzed muscle and use it for controlling the stimulation of the same muscle.

Non-linear short-term prediction in speech coding

Addresses the question of how to extract the nonlinearities in speech with the prime purpose of facilitating coding of the residual signal in residual excited coders. The short-term prediction of speech in speech coders is extensively based on linear models, e.g. the linear predictive coding technique (LPC), which is one of the most basic elements in modern speech coders. This technique does not allow extraction of nonlinear dependencies. If nonlinearities are absent from speech the technique is sufficient, but if the speech contains nonlinearities the technique is inadequate. The authors give evidence for nonlinearities in speech and propose nonlinear short-term predictors that can substitute the LPC technique. The technique, called nonlinear predictive coding, is shown to be superior to the LPC technique. Two different nonlinear predictors are presented. The first is based on a second-order Volterra filter, and the second is based on a time delay neural network. The latter is shown to be the more suitable for speech coding applications.<<ETX>>

Experimental comparison of signal subspace based noise reduction methods

The signal subspace approach for non-parametric speech enhancement is considered. Several algorithms have been proposed in the literature but only partly analyzed. Here, the different algorithms are compared, and the emphasis is put onto the limiting factors and practical behavior of the estimators. Experimental results show that the signal subspace approach may lead to a significant enhancement of the signal to noise ratio of the output signal.

Objective speech quality assessment of compounded digital telecommunication systems

Digital telecommunication networks involve multiple number of public switched telephone networks (PSTN), cellular and mobile systems and to some extent also satellite systems. Most of these networks contain non-linear speech coders which may degrade the overall end-to-end quality of speech. An important problem is how to assess the speech quality of such compounded systems. The object of this paper is to describe the development of a proposed three-layer model for objective speech quality assessment. The design of the objective measure is based on a subjective test of the speech quality of 16 compounded transmission paths and involves 40 subjects using 21 different rating scales. The subjective test supports the definition of four main perceptual dimensions to be used in the second layer of the proposed model. These defined perceptual elements form the basis of designing an objective measure for predicting the total quality of a given telecommunication connection.

Quantization of non-linear predictors in speech coding

Discusses how to exploit the nonlinearities in speech with the main purpose of improving the prediction in speech coders. If non-linearities are absent from speech the linear technique is sufficient, but if non-linearities are present the technique is inadequate and more sophisticated predictors are called for. Thyssen et al. (1994) gave evidence for non-linearities in speech and presented two non-linear short-term predictors that both were superior to the linear predictor without quantization. The present authors give methods to design vector quantizers for the non-linear predictors and investigate how vector quantization of the non-linear predictors affects prediction. Furthermore, they compare the performance of the quantized non-linear predictors to the performance of traditional quantized linear predictors. The experiments show that 10-bit VQ of the non-linear predictor leads to similar performance as 20-bit state-of-the-art split VQ of the LSP-parameters.

Speech quality assessment of compounded digital telecommunication systems; perceptual dimensions

Digital telecommunication networks may involve a multiple number of public switched telephone networks (PSTN), cellular and mobile systems and to some extent also satellite systems. Most of these networks contain non-linear speech coders and other speech algorithms which may degrade the overall end-to-end quality of speech. An important problem is how to assess the speech quality of such compounded systems. The object of this paper is to describe the first stage of the construction of a proposed three-layer model for speech quality assessment. A subjective test of the speech quality of 16 different compounded transmission paths (mixtures of PCM, GSM full and half rate, DECT, CELP, LD CELP, FS10-16) is carried out by 40 subjects using 21 different rating scales. The main result of this paper is the test results which lead to the definition of four main perceptual dimensions to be used in the second layer of the proposed model.

Modeling and evaluation of multimodal perceptual quality

The increasing performance requirements of multimedia modalities, carrying speech, audio, video, image, and graphics emphasize the need for assessment methods of the total quality of a multimedia system and methods for simultaneous analysis of the system components. It is important to take into account still more perceptual characteristics of the human auditory, visual, tactile systems, as well as combinations of these systems, it is also highly desirable to acquire methods for analysing the main perceptual parameters, which constitute the input for the total quality assessment. A framework is suggested for assessing the quality of modalities and their combinations.

Using neural networks for vector quantization in low rate speech coders

The problem of reducing the complexity of the codebook search in low-rate speech coders is addressed. Emphasis is placed on vector quantization of the short-term parameters (spectral parameters), where the increasing demand for higher performance necessitates codebook sizes of approximately 2/sup 20/. As full search is impractical, a novel path search algorithm is proposed. it is based on a multidimensional version of Kohonens self-organizing feature map, using the ordering aspects of the map. A comparison with the full-search LBG algorithm shows a substantial reduction in search complexity with only a minor degradation in speech quality. Furthermore, the speech quality is better than that obtained with split-LBG.<<ETX>>

Reduction of general broad-band noise in speech by truncated QSVD: implementation aspects

Publisher Summary In many speech processing applications an appropriate filter is needed to remove noise. The truncated singular value decomposition (SVD) technique has a noise filtering effect and, provided that the noise is white, it can be applied directly in noise reduction algorithms. However, for non-white broad-band noise a pre-whitening operation is necessary. This chapter focuses on implementation aspects of a newly proposed quotient singular value decomposition (QSVD)-based algorithm for reduction of general broad-band noise in speech. A distinctive advantage of the algorithm is that the pre-whitening operation is an integral part of the algorithm, and this is essential in connection with updating issues in real-time applications. The chapter compares the existing implementation (based on the QSVD) with an implementation based on the ULLV decomposition that can be updated at a low cost.