Search

Some Speech Enhancement algorithms based on the iterative Wiener filtering Method due to L1m-Oppenheim [2] are presented. In the original Lim-Oppenheim algorithm, speech AR estimation is carried out using classic second-order analysis, but our algorithms consider a more robust AR modelling. Two different strategies of speech AR estimation are presented and both estimators are trying to see as less amount of noise as possible. First one uses a previous One-Sided Autocorrelation computation, that is a pole-preserving function, and the actual SNR m the second-order LPC analysis is increased. Second one combines advantages of Higher-Order Statistics [1] with a linear combination of AR coefficients, belonging to two consecutive overlapped frames, to assess a less disturbed speech estimation.

In this paper a phoneme recognition system based on predictive neural networks is proposed. Neural networks are used to predict observation vectors of speech frames. The obtained prediction error is used for phoneme recognition as 1) distortion measure on the frame level and 2) as feature, which is statistically modeled by the Rayleigh distribution. Continuous speech phoneme recognition experiments are performed different settings of the system are evaluated.

In previous works [5], [6], we studied some speech enhancement algorithms based on the iterative Wiener filtering method due to Lim-Oppenheim [2], where the AR spectral estimation of the speech is carried out using a second-order analysis. But in our algorithms we consider an AR estimation by means of cumulant analysis. This work extends some preceding papers due to the authors: a cumulant-based Wiener Filtering (AR3_IF) is applied to Robust Speech Recognition. A low complexity approach of this algorithm is tested in presence of bathroom water noise and its performance is compared to classical Spectral Subtraction method. Some results are presented when training task of the speech recognition system (HTK-MFCC) is executed under clean and noisy conditions. These results show a lower sensitivity to the presence of water noise when applying AR3_IF algorithm inside of a speech recognition task.

This paper describes a coding scheme for broadband speech (sampling frequency 16KHz). We present a wideband speech encoder called APVQ (Adaptive Predictive Vector Quantization). It combines Subband Coding, Vector Quantization and Adaptive Prediction as it is represented in Fig. I. Speech signal is split in 16 subbands by means of a QMF filter bank and so every subband is 500Hz wide. This APVQ encoder can be seen as a vectorial extension of a conventional ADPCM encoder. In this scheme, signal vector is formed with one sample of the normalized prediction error signal coming from different subbands and then it is vector quantized. Prediction error signal is normalized by its gain and normalized prediction error signal is the input of the VQ and therefore an adaptive Gain-Shape VQ is considered. This APVQ Encoder combines the advantages of Scalar Prediction and those of Vector Quantization. We evaluate wideband speech coding in the range from 1 to 2 bits/sample, that leads to a coding rate from 16 to 32 kbps.

We study some speech enhancement algorithms based on the iterative Wiener filtering method due to Lim-Oppenheim [2], where the AR spectral estimation of the speech is carried out using a second-order analysis. But in our algorithms we consider an AR estimation by means of cumulant analysis. This work extends some preceding papers due to the authors, where information of previous speech frames is taken to initiate speech AR modelling of the current frame. Two parameters are introduced to dessign Wiener filter at first iteration of this iterative algorithm. These parameters are the Interframe Factor (IF) and the Previous Frame Iteration (PFI). A detailed study of them shows they allow a very important noise suppression after processing only first iteration of this algorithm, without any appreciable increase of distortion. Finally, the simplest cumulant-based algorithm is applied to Speech Recognition and some preliminary results are presented.

Some Speech Enhancement algorithms based on the iterative Wiener filtering Method due to LimOppenheim [2] are presented. In the original Lim-Oppenheim algorithm, AR spectral estimation of speech is carried out using a second-order analysis, but our algorithms consider an AR estimation by means of cumulant analysis. This work extends some preceding papers due to the authors. Information of previous speech frames is taken to initiate speech AR modeling of the current frame and, so, two parameters are introduced to dessign Wiener Filter at first iteration of every frame. Another algorithm obtains speech AR estimation in the autocorrelation domain. Both algorithms are compared to classical second-order algorithm (AR2) and third-onler cumulant-based algorithm (AR3), when car noise disturbs clean speech signal. A detailed study shows that boths techniques significantly increase noise suppression after first iteration processing and, therefore, convergence speed of this iterative algorithm is strongly accelerated.

This paper describes a coding scheme for broadband speech. It can be seen as a vectorial extension of an conventional ADPCM encoder. In this scheme, the vector signal is formed with one sample of the normalizaed prediction error of each subband and then, it is vector quantized. It combines the advantages of the scalar prediction and of the vector quantization (VQ) . We handle the high vector dimensionality by using a multi-VQ. It requires a previous subvector division and an adequate bit assignement among them. This scheme shows an high capacity to drive large dynamic range signals like broadband speech.

Recently, a new parametrization technique based on the AR modelling of the one-sided autocorrelation sequence (OSALPC) has shown to be attractive for speech recognition because of its simplicity and its high recognition perfomance in noisy conditions. In this paper, that new parametrization technique is proposed to speaker identification in noisy enviroment. Experimental results obtained with a new speaker identification system based on the statistics of the cepstrals vectors show that OSALPC also achieves much better results than standard parametrization techniques.

This paper describes a coding scheme for broadband speech (sampling frequency 16KHz). We present a wideband speech encoder called APVQ (Adaptive Predictive Vector Quantization). It combines Subband Coding, Vector Quantization and Adaptive Prediction as it is represented in Fig.1. Speech signal is split in 16 subbands by means of a QMF filter bank and so every subband is 500Hz wide. This APVQ encoder can be seen either as a vectorial extension of a conventional ADPCM encoder or as a scalar Subband AVPC encoder [1],[3]. In this scheme, signal vector is formed with one sample of the normalized prediction error signal coming from different subbands and then it is vector quantized. Prediction error signal is normalized by its gain and normalized prediction error signal is the input of the VQ and therefore an adaptive Gain-Shape VQ is considered. This APVQ Encoder combines the advantages of Scalar Prediction and those of Vector Quantization. We evaluate wideband speech coding in the range from 1 to 2 bits/sample.

Recently, 4G mobile phone systems have been designed to process wideband speech signals whose sampling frequency is 16 kHz. However, most part of mobile and classical phone network, and current 3G mobile phones, still process narrowband speech signals whose sampling frequency is 8 kHz. During next future, all these systems must be living together. Therefore, sometimes a wideband speech signal (with a bandwidth up to 7,2 kHz) should be estimated from an available narrowband one (whose frequency band is 300-3400 Hz). In this work, different techniques of audio bandwidth extension have been implemented and evaluated. First, a simple non-model-based algorithm (interpolation algorithm) has been implemented. Second, a model-based algorithm (linear mapping) have been designed and evaluated in comparison to previous one. Several CMOS (Comparison Mean Opinion Score) [6] listening tests show that performance of Linear Mapping algorithm clearly overcomes the other one. Results of these tests are very close to those corresponding to original wideband speech signal.

This paper describes a coding scheme for broadband speech. It can be seen as a vectorial extension of a conventional ADPCM encoder. In this scheme, signal vector is formed with one sample of the normalized prediction error of each subband and then it is vector quantized. It combines the advantages of the scalar prediction and those of vector quantization (VQ). We handle the high vector dimensionality by using a multiVQ. It requires a previous subvector division and an adequate bit assignment among them. This scheme shows a high capacity to drive large dynamic range signals like broadband speech. Predictor and codebook dessigns are discussed. Some results about speech prediction and coding are reported.

We study some speech enhancement algorithms based on the iterative Wiener filtering method due to Lim-Oppenheim [2], where the AR spectral estimation of the speech is carried out using a second-order analysis. But in our algorithms we consider an AR estimation by means of cumulant analysis. This work extends some preceding papers due to the authors, providing a behavior comparison between cumulant algorithms and classical auto-correlation one. Some results are presented considering AWGN that allows the best improvement and those noises (diesel engine and reactor noises) that leads to the worst one. An exhaustive empirical test shows that cumulant algorithms outperform the original autocorrelation algorithm, specially at low SNR.