DSP-based oversampling adaptive noise canceller for background noise reduction for mobile phones

In this paper, a DSP-Based of an oversampling adaptive noise canceller for background noise reduction using the Recursive Least Squares (RLS) algorithm on a Digital Signal Processor (DSP) is presented and its performance discussed. The implementation described seeks to improve the signal to background noise ratio for effective telephone communication in adverse non-stationary acoustic environments. Real-time experiments oversampling at 96 kHz were carried out on a DSP board (Texas Instruments TMS320C6713 DSK). Primary input signal to noise ratios from 0 dB to 10 dB were proposed for the cancelling performance tests. Results were analyzed in time and frequency domain. Also, a practical method based on the normalized performance function to objectively assess the adaptive noise canceller performance is presented. Mean Opinion Score (MOS) listening test along three dimensions following the ITU-T P.835 procedures were applied to subjectively assess the adaptive noise canceller output quality.