laughlin@fornax.UUCP (Bob Laughlin) (02/14/91)
Hello, I am looking for information on algorithms that can be applied to digitized sound data that yield psychacoustically relevant measures of "average sound level" and "dynamic range". These measures would be computed at specified intervals over the duration of the sound (say at 200 ms intervals). "Average sound level" should yield a measure of the overall sound intensity over a given interval and "dynamic range" should yield some measure of contrast between loud and soft sounds over some duration. If anybody knows of relevant literature on this I'd appreciate it if you emailed it to me or posted it here. Thank you. -- Bob Laughlin laughlin@cs.sfu.ca
sandell@batcomputer.tn.cornell.edu (Gregory Sandell) (02/15/91)
In article <2112@fornax.UUCP> laughlin@fornax.UUCP (Bob Laughlin) writes: > > Hello, > > I am looking for information on algorithms that can be applied >to digitized sound data that yield psychacoustically relevant >measures of "average sound level" and "dynamic range". >These measures would be computed at specified intervals >over the duration of the sound (say at 200 ms intervals). >"Average sound level" should yield a measure of the overall >sound intensity over a given interval and "dynamic range" >should yield some measure of contrast between loud and >soft sounds over some duration. If anybody knows of relevant >literature on this I'd appreciate it if you emailed it to me >or posted it here. Thank you. >-- DIGITAL SIGNAL PROCESSING OF SPEECH SIGNALS by Rabiner and Schafer provide some good material on calculating sound level from samples. A real simple measure is RMS energy. Over the duration interval you want to measure, square each sample (to get rid of the sign) and keep a running sum. Take the mean of that sum, and take its square root. You might want to convert this measure to dB power. Assuming we are talking about 16-bit samples, you could try using this: 10.0 * log10(amp/32767.0) This will represent the most intense signal as 0 dB and all other levels as negative. I can give you some references to classic literature on loudness calculation. However, much of that will be included in the bibliographies of Rabiner & Schafer. Let me know if you need more. Greg Sandell