| Paper: | SLP-P18.10 | ||
| Session: | LVCSR Systems | ||
| Time: | Friday, May 19, 10:00 - 12:00 | ||
| Presentation: | Poster | ||
| Topic: | Speech and Spoken Language Processing: Decoding algorithms and implementation | ||
| Title: | A Decoder For LVCSR Based On Fixed-Point Arithmetic | ||
| Authors: | Enrico Bocchieri, Doug Blewett, AT&T Labs – Research, United States | ||
| Abstract: | The increasing computational power of embedded CPU's motivates the fixed-point implementation of highly accurate large-vocabulary continuous-speech (LVCSR) algorithms, to achieve the same performance on the device as on the server. We report on methods for the fixed-point implementation of the frame-synchronous beam-search Viterbi decoder, HMM likelihood computation, and N-grams language models. Our fixed-point recognizer is as accurate as the floating-point recognizer in several LVCSR experiments on the DARPA Switchboard and on an AT&T proprietary task. We also present results on the DARPA Resource Management task using the 206 MHz StrongARM-1100 CPU, where the fixed-point implementation enables real-time performance: the floating-point recognizer (with floating-point software emulation) is 50 times slower, for the same accuracy. | ||
©2018 Conference Management Services, Inc. -||- email: webmaster@icassp2006.org -||- Last updated Friday, August 17, 2012