Automatic speech processing

access password

16.ASP+lec9$11

In this part of lecture, we saw how two speech utterances can be matched in a similar manner as matching of two strings. We saw two cases: (a) speech signals are converted into sequence of phones and are matched through dynamic programming and (b) speech signal is converted into sequence of acoustic feature vectors and are matched through dynamic programming. At the end, we saw how this kind of matching can be used to select the best matching speech signal from a set of speech signals given a speech signal as input (foundation for automatic speech recognition)