Digital signal processing in studies of animal acoustical communication, including human speech

The study of animal acoustical communication and other biological signals like cardiopulmonary cycles and neurological waves has advanced rapidly in the last few years. Progress has been stimulated by dramatic improvements in transducers, recorders, computers, and digital signal processing (DSP) techniques. Nevertheless, many biologists cannot take advantage of these new techniques and technology because mainframe computer usage is expensive and DSP theory is buried in highly technical engineering and mathematical journals. Recently, new developments in microcomputer hardware and commercially available DSP software have greatly improved this picture. Nevertheless, DSP theory remains largely inaccessible to the non-specialist. In this paper, selected areas of DSP theory are surveyed in relation to a common biological signal processing problem, namely, the analysis of animal acoustical signals. Signal digitization techniques are described, including the concepts of word-width, sample rate, and data encoding. An overview of the mathematical models developed for human speech encoding and analysis is presented. Finally, selected DSP techniques used in the time and frequency domains, including digital filtering, spectral analysis, and dynamic time warping are reviewed.