Title :
A 0.9V, 30 µ W feature extractor for remote speech recognition
Author :
Ferrari, A. ; Borgatti, M. ; Felici, M. ; Guerrieri, R.
Author_Institution :
DEIS - Università di Bologna, Italy
Abstract :
A feature extraction chip for speech recognition computes fifteen cepstra each 8ms at 64kHz clock rate and dissipates 30µW at 0.9V. It has been implemented as a gate array in a 0.5µm, three-metal CMOS technology. The average energy required to process a single word of the TI46 speech corpora is 10µJ. It achieves recognition rates over 98% in isolated-word, speech recognition tasks.
Keywords :
Cascading style sheets; Cepstral analysis; Channel coding; Data mining; Feature extraction; Speech analysis; Speech coding; Speech processing; Speech recognition; Voltage;
Conference_Titel :
Solid-State Circuits Conference, 1997. ESSCIRC '97. Proceedings of the 23rd European
Conference_Location :
Southampton, UK
