嵌入式语音识别中的卡尔曼滤波及优化

Optimization and Application of Kalman Filter in the Embedded Speech Recognition System

嵌入式语音识别系统的性能在环境噪音下往往受到严重影响。为了消除噪音对系统识别性能的影响,在嵌入式系统平台中引入了卡尔曼滤波方法,结合嵌入式平台的限制及系统实时性的要求,对卡尔曼滤波算法进行了优化和调整,提出了基于FPGA的卡尔曼滤波精简算法。实验证明在减少计算量和降低计算精度的条件下,该方法仍有效地消除了环境噪音,提高了语音信号的信噪比。

In nosy environment, the embedded-speech-recognize system usually got bad performance.Kalman filtering method was introduced to embedded system platform for the above-mentioned problem.Combining embedded platform limits and the requirement of real-time system, an improved algorithm, which was based on traditional Kalman filter algorithm, was proposed for FPGA.The experiment results show the algorithm can not only reduce the calculation and precision, but also can eliminate environment noise and improve SNR.