Abstract:

A recent analysis of moving sound sources has revealed that information in the dynamic acoustic array can specify fundamental attributes of the source, such as its distance, speed, and spectrum [R. L. Jenison, Ecol. Psych. (1997)]. A model based upon the Kalman filter is introduced which successfully estimates such distal properties using only observed proximal information. The experiment investigated whether 3 specific forms of acoustic information (amplitude, frequency, interaural phase differences) are necessary and sufficient to estimate the intrinsic frequency of a simulated moving sound source. Simulations show the model most accurately estimates intrinsic frequency when all three sources of acoustic information are included in the signal; when dynamic changes in one or more of these parameters are removed, the model's performance degrades. To test human subjects, a technique was developed to generate moving signals from a sample of a natural moving sound source. This process allowed for precise control over the signal's virtual and acoustic parameters. Subjects in a 2AFC design are required to guess the unshifted spectra of the moving stimuli as dynamic changes in the acoustic parameters are systematically removed. [Work supported by NSF.]