基于FxLMS算法和预测滤波器的数字耳机降噪研究

Digital Headphone Noise Cancellation Research Based on FxLMS Algorithm and Prediction Filter

通常数字降噪耳机的自适应滤波器采用最小均方差(LMS)、归一化最小均方差(NLMS)等算法,但是由于降噪耳机电路中存在次级通道延迟和AD/DA转换延迟,导致这些滤波器输出信号无法与噪声信号协调起来。FxLMS算法能够有效补偿次级通道延迟,同时在FxLMS滤波器基础上,增加预测滤波器用于补偿AD/DA转换延迟。仿真中,用增加预测滤波器的FxLMS滤波器对已采样的发动机噪声进行降噪,并将其降噪效果与LMS自适应滤波器的降噪效果进行比较。结果表明,在处理低频噪声时,改进后的FxLMS滤波器降噪效果优于LMS自适应滤波器。

The adaptive filter based on LMS and NLMS algorithms is usually used to remove the noise in digital headphones. Because of the latency of secondary-path and AD/DA conversion, these filter output signal can＇t collaborate with the noise. FxLMS algorithm can compensate for secondary-path delay effec- tively. We combined the prediction filter with FxLMS filter to compensate for AD/DA conversion delays. In the simulation,the FxLMS filter with prediction filter was used to reduce the sampled noise of engine and the noise cancellation effect was compared with LMS adaptive filter. The results indicate that the improved FxLMS filter is superior to the LMS adaptive filter.