This work presents a general overview of a new speech processing and recognition tool (SPEAR). The SPEAR system handles speech signals in different analysis levels. First, a signal processing stage is used to adjust and analyze the speech waveform with different techniques (Fourier analysis, noise filtering, signal segmentation, etc). Second, a feature extraction and data mining module are used to extract the main parameters needed for describing every speech segment (formant frequencies, energy, LP, cepstrum and PLP coefficients, etc.). The speech recognition is performed in two phases. The first one is a frame by frame phone estimation by means of a set of parallel associative memories and the second one uses a DTW algorithm for complete word matching. Experimental results show a good performance of the SPEAR tool for speaker dependent and independent tests under noisy environments (over 15 dB S/N rate).
General Overview of the SPEAR Speech Recognition System
CURATELLI, FRANCESCO;
1999-01-01
Abstract
This work presents a general overview of a new speech processing and recognition tool (SPEAR). The SPEAR system handles speech signals in different analysis levels. First, a signal processing stage is used to adjust and analyze the speech waveform with different techniques (Fourier analysis, noise filtering, signal segmentation, etc). Second, a feature extraction and data mining module are used to extract the main parameters needed for describing every speech segment (formant frequencies, energy, LP, cepstrum and PLP coefficients, etc.). The speech recognition is performed in two phases. The first one is a frame by frame phone estimation by means of a set of parallel associative memories and the second one uses a DTW algorithm for complete word matching. Experimental results show a good performance of the SPEAR tool for speaker dependent and independent tests under noisy environments (over 15 dB S/N rate).
I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
