Robust Speech Recognation and Understanding by Eric K. Ringger

By Eric K. Ringger

Show description

Read Online or Download Robust Speech Recognation and Understanding PDF

Best physics books

Halogen oxides- radicals, sources and reservoirs in the laboratory and in the atmosphere

Learn job in atmospheric chemistry has persisted to speed up in recent times, and there's now heightened public understanding of the environmental matters during which it performs an element. This ebook appears to be like on the new insights and interpretations afforded by means of the hot advances, and areas in context those advancements.

Econophysics of Markets and Business Networks

Econophysicists have lately been particularly winning in modelling and analysing numerous monetary structures like buying and selling, banking, inventory and different markets. The statistical behaviour of the underlying networks in those structures have additionally been pointed out and characterized lately. This ebook experiences the present econophysics researches within the constitution and functioning of those advanced monetary community structures.

Extra info for Robust Speech Recognation and Understanding

Sample text

These results also indicate the importance of speech detection in speaker-clustering procedures. Figure 8. Overall speaker-tracking results plotted with DET curves. Lower DET values correspond to better performance. The overall performance of the evaluated speaker-diarisation (SD) system is depicted in Figure 8, where the overall speaker-tracking results are shown.

It is referred to as the EN-phones recognizer in all our experiments. Both phoneme recognizers were constructed from the HMMs of monophone units joined in a fully connected network. , 2004). The phoneme sets of each language were different. In the SI-phones recognizer, 38 monophone base units were used, while in the TIMIT case, the base units were reduced to 48 monophones, according to (Lee & Hon, 1989). In both recognizers we used bigram phoneme language models in the recognition process. The recognizers were also tested on parts of the training databases.

In both cases they used these features for speech/music classification, but the idea could be easily extended to the detection of speech and non-speech signals, in general. The basic motivation in both cases was to obtain and use features that were more robust to different kinds of music data and at the same time perform well on speech data. , 2006a). In this way we also examined the behaviour of a phoneme recognizer, but the functioning of the recognizer was measured at the output of the recognizer rather than in the inner states of such a recognition engine.

Download PDF sample

Rated 4.07 of 5 – based on 50 votes