If the audio signal varies on one dimension only (such as frequency), the number of signals to be identified should not exceed four.