Accurate endpoint detection is important for improving the speech recognition capability. This paper proposes a novel endpoint detection method which combines energy-based and likelihood ratio-based voice activity detection (VAD) criteria, where the likelihood ratio is calculated with speech/non-speech Gaussian mixture models (GMMs). Moreover, the proposed method introduces the discriminative feature extraction technique (DFE) in order to improve the speech/non-speech classification. The DFE is used in the training of parameters required for calculating the likelihood ratio. Experimental results have shown that the proposed endpointer achieves good performance compared to an energy-based endpointer in terms of start-of-speech (SOS) and end-of-speech (EOS) detections. Due to the improvement of the endpointer, the performance of automatic speech recognition (ASR) has also been improved.

The major drawback of most noise reduction methods in speech applications is the annoying residual noise known as musical noise. A potential solution to this artifact is the incorporation of a human hearing model in the suppression filter design. However, since the available models are usually developed in the frequency domain, it is not clear how they can be applied in the signal subspace approach for speech enhancement. In this paper, we present a Frequency to Eigendomain Transformation (FET) which permits to calculate a perceptually based eigenfilter. This filter yields an improved result where better shaping of the residual noise, from a perceptual perspective, is achieved. The proposed method can also be used with the general case of colored noise. Spectrogram illustrations and listening test results are given to show the superiority of the proposed method over the conventional signal subspace approach.

The major drawback of most noise reduction methods is what is known as musical noise. To cope with this problem, the masking properties of the human ear were used in the spectral subtraction methods. However, no similar approach is available for the signal subspace based methods. In a previous work, we presented a frequency to eigendomain transformation which provides a way to calculate a perceptually based upper bound for the residual noise. This bound, when used in the signal subspace approach, yields an improved result where better shaping of the residual noise is achieved. In this paper, we further improve this method and provide an easy way to generalize it to the colored noise case. Listening tests results are given to show the superiority of the proposed method.

We extend the single microphone signal subspace approach for speech enhancement, to a multi-microphone design. In the single microphone case, the tradeoff between speech quality and intelligibility is an handicap which limits its performance. This is because it is based on a linear speech model which does not usually offer enough degrees of freedom for noise reduction. In our method, we show how we can easily, and with comparable computational complexity, get more degrees, of freedom by using signals from more than one microphone. Experimental results show that this leads to improvements in the noise reduction performance.

In this paper we present a subband scheme to decrease the computational complexity of simulating room responses to acoustic signals, important for instance in microphone array systems. Besides, the new method offers added flexibility to the well known image method by allowing to choose the reflection coefficients of every frequency subband independently of each others. The efficiency of this method is tested experimentally.

In this paper, a new set of speech feature parameters based on multirate signal processing and the Teager energy operator is developed. The speech signal is first divided into nonuniform subbands in mel-scale using a multirate filter-bank, then the Teager energies of the subsignals are estimated. Finally, the feature vector is constructed by log-compression and inverse DCT computation. The new feature parameters have a robust speech recognition performance in car engine noise which is low pass in nature.

In this letter, a new set of speech feature parameters based on multirate signal processing and the Teager energy operator is introduced. The speech signal is first divided into nonuniform subbands in mel-scale using a multirate filterbank, then the Teager energies of the subsignals are estimated. Finally, the feature vector is constructed by log-compression and inverse discrete cosine transform (DCT) computation. The new feature parameters have robust speech recognition performance in the presence of car engine noise.