In this paper we present an algorithm for automatic classification of sound into speech, instrumental sound/ music and silence. The method is based on thresholding of features derived from the modulation envelope of the frequency limited audio signal. Four characteristics are examined for discrimination: the occurrence and duration of energy peaks, rhythmic content and the level of harmonic content. The proposed algorithm allows classification directly on MPEG-1 audio bitstreams. The performance of the classifier was evaluated on TRECVID test data. The test results are above-average among all TREC participants. The approaches adopted by other research groups participating in TREC are also discussed.