We augment the Mel cepstral (MFCC) feature representation with voicing features from an independent front end. The voicing feature front end parameters are optimized for recognition accuracy. The voicing features computed are the normalized autocorrelation peak and a newly proposed entropy of the high-order cepstrum. We explored several alternatives to integrate the voicing features into SRI’s DECIPHER system. Promising early results were obtained in a simple system concatenating the voicing features with MFCC features and optimizing the voicing feature window duration. Best results overall came from a more complex system combining a multiframe voicing feature window with the MFCC plus third differential features using linear discriminant analysis and optimizing the number of voicing feature frames. The best integration approach from the single-pass system experiments was implemented in a multi-pass system for large vocabulary testing on the Switchboard database. An average WER reduction of 2 % relative was obtained on the NIST Hub-5 dev2001 and eval2002 databases. 1.

Automatic audio classification usually considers sounds as music, speech, silence or noise, but works about the noise class are rare. Audio features are generally specific to speech or music signals. In this paper, we present a new audio feature sets that lead to the definition of four classes: colored, pseudo-periodic, impulsive and sinusoids within noises. This classification relies on works about the perception of noises. This audio feature set is experimented for noisy sound segmentation. Noise-to-noise transitions are characterized by means of statistical decision model based on Bayesian framework. This statistical method has been trained and experimented both on synthetic and real audio corpus. Using proposed feature set increases the discriminant power of Bayesian decision approach compared to a usual feature set. 1.

We announce the public availability of the RWTH Aachen University speech recognition toolkit. The toolkit includes state of the art speech recognition technology for acoustic model training and decoding. Speaker adaptation, speaker adaptive training, unsupervised training, a finite state automata library, and an efficient tree search decoder are notable components. Comprehensive documentation, example setups for training and recognition, and a tutorial are provided to support newcomers. Index Terms: speech recognition, LVCSR, software 1.

Spectutils is a GNU Octave toolkit for analyzing and visualizing audio signals. Spectutils allows to display oscillograms, FFT spectrograms as well as pitch detection graphs. Spectutils can best be characterized as a user interface for GNU Octave, which integrates signal analysis and visualization functionality into dedicated function calls. Therefore, signal analysis with Spectutils requires little or no prior knowledge of Octave or MATLAB programming. 1.