Improve detection performance of speech recognizer in an automotive environment

Abstract

In-car speech recognition is a challenging area of research. The area has been studied, and more of is concentrated on addressing the issues in acoustic echo cancellation (AEC). In application such as a voice controlled car audio system, voice commands by the driver are corrupted by audio out of loudspeaker. In this paper we propose and implement a robust, low complex voice controller for car audio system involving an efficient AEC and an effective detection module for a simple speech recognizer The proposed method has been implemented on OMAP 2420 TI-DSP C55x, with a performance of under 59 mega cycles for complete system and is tested in real time environment.