Recent research has shown that speech can be sparsely represented using a dictionary of speech segments spanning multiple frames, exemplars, and that such a sparse representation can be recovered using Compressed Sensing techniques. In previous work we proposed a novel method for noise robust automatic speech recognition in which we modelled noisy speech as a sparse linear combination of speech...

There is an enormous amount of multi-media data including images, video, speech, audio and text, distributed among the various computer nodes on the Internet. The extent to which a user wiU be able to derive useful information from these data depends largely on the ease with which required data can be retrieved from the databases. The share volume of the data also poses a storage constraint on ...

Prevalent speaker recognition methods use only spectralenvelope based features such as MFCC, ignoring the rich speaker identity information contained in the temporalspectral dynamics of the entire speech signal. We propose a new feature called compressed spectral dynamics or CSD for speaker recognition based on a compressed time-frequency representations of spoken passwords which effectively ca...

In current speech recognition systems, speech is represented by a 2-D sequence of parameters that model the temporal evolution of the spectral envelope of speech. Linear transformation or filtering along both time and frequency axes of that 2-D sequence are used to enhance the discriminative ability and robustness of speech parameters in the HMM pattern-matching formalism. In this paper, we com...

SPEECH Hynek Hermansky, Carlos Avendano and Eric A. Wan Oregon Graduate Institute of Science & Technology Department of Electrical Engineering and Applied Physics P.O. Box 91000, Portland, OR 97291 ABSTRACT Finite Impulse Response (FIR) Wiener-like lters designed on training data are applied to time trajectories of cubic-root compressed short-term power spectrum of noisy speech recorded over ce...

In this paper the question is addressed whether the word-level intelligibility of time-compressed speech can be improved over linear compression by using a type of non-linear compression. Two options are tested: one type of compression which takes into account the natural timing of fast speech; and one other type of compression that saves the segmental intelligibility of short unstressed syllab...

This paper presents an experimental implementation of a low−complexity speaker recognition algorithm working in the compressed speech domain. The goal is to perform speaker modeling and identification without the need of decoding the speech bitstream to extract speaker dependent features, thus saving important system resources, for instance, in mobile battery powered DSP devices. Bitstream valu...

we give some new results on sparse signal recovery in the presence of noise, forweighted spaces. traditionally, were used dictionaries that have the norm equal to 1, but, forrandom dictionaries this condition is rarely satised. moreover, we give better estimationsthen the ones given recently by cai, wang and xu.