New time-domain analysis for complex animal vocalizations

Abstract

The time-domain analysis introduced here incorporates a transformation of the acoustic waveform, which is functionally analagous to the physiological operation of phase-locking in the peripheral auditory nervous system. The acoustic signals are sampled and digitized at 20 kHz. Then the time of up-crossing is determined for each cycle in the waveform. Each cycle is individually evaluated with regard to three classes of time-domain parameters, namely, cycle frequency, cycle amplitude, and cycle microstructure. Some of this information is graphically presented by a Log Inverse Period (LIP) plot, a modification of the “instantaneous frequency” plot used by neurophysiologists. This plot is particularly useful for cleanly delineating the fine temporal structure and acoustic transients embedded in complex waveforms. The chief advantage of applying these time-domain techniques is their perfect time resolution. In contrast to the classical frequency-domain analyses, these time-domain analyses have no inherent bandwidth limitations. This fact is significant when the acoustic signal to be analyzed contains rapidly changing acoustic states believed to bear information. Especially where the auditory temporal resolution of humans may not be equal to that of the animals which communicate with complex vocalizations, precise temporal resolution is desirable. Analysis results are shown with examples drawn largely from avian vocalizations.