A Low-Power, Fixed-Point, Front-End Feature Extraction for a Distributed Speech Recognition System