Hi All I want to test the performance of a noise suppression algorithm running on a 5402 DSK . Is it appropriate to use a known SNR noise corrupted file as an input to the codec and record the output of the codec on a pc (using something like a wav recorder) to make such measurements. I am not sure, how much will things like computer fan noise/ board noise etc. might downplay on the performance evaluation. The wav recorder etc. seem to be very sensitive to recording/playback volume levels etc. which adds normalization issues too - if i am not mistaken... What would be a better way to do a performance evaluation in a "more real time" manner, instead of simply feeding audio files in form of data files using probe points. Would appreciate any comments or careabouts to take into consideration the complete hardware loop , but yet not undermining the performance of the core code itself. Rgds Mukul __________________________________________________ |
SNR measurements
Started by ●September 17, 2002