I have a few thousand small 2MB audio recordings from a recording device that turns over every so often. The total amount of recording time is several days, but each file contains about 5 minutes worth. Normally, it is only meant to record audio when there is some volume happening (speech). Problem is that I had the sensitivity setting too sensitive. The result is a mix of audio files with speech in it, but with the bulk of them only having recorded a soft background hiss. I can't just "grep" the files for a bunch of zeros. I was thinking of writing a quick little Java application that loads each file, and then calculate the amount of energy in it, by simply adding the different between each sample. But I was hoping that someone might now a simple clever way of wiring together some bash/awk/play/mpg123 something anything. Possibly there exists a command line app in one of the audio tool sets that I'm not aware of yet?