Average Reviews:
(More customer reviews)This book's most interesting sections describe the state of the art in 1999 of automatic speech recognition. The dominance of Hidden Markov Models and Linear Predictive Coding had been clear by then, for practical, near-realtime ASRs.
There is reference in several sections to the high speed microprocessors and cheap memory and disk space that made all this possible. Given that this is 2004 and the book came out in 1999, it means that in the described research, typically the CPU clocks were under 1 GHz, and memory was maybe less than 100 Mb. In other words, with current hardware, you should be able to implement most of the described algorithms with possibly greater efficacy than achieved in the text. The other implication is that now there might be greater commercial applications possible using these methods.
There is some discussion of how to improve ASR recognition. But this can run into the currently intractable problem of trying to infer semantics from an arbitrary utterance. Hard AI.
Click Here to see more reviews about: 1999 IEEE International Conference on Acoustics, Speech and Signal Processing (Icassp)
No comments:
Post a Comment