Keith D. Martin

Musical instrument identification: A pattern‐recognition approach

A statistical pattern‐recognition technique was applied to the classification of musical instrument tones within a taxonomic hierarchy. Perceptually salient acoustic features—related to the physical properties of source excitation and resonance structure—were measured from the output of an auditory model (the log‐lag correlogram) for 1023 isolated tones over the full pitch ranges of 15 orchestral instruments. The data set included examples from the string (bowed and plucked), woodwind (single, double, and air reed), and brass families. Using 70%/30% splits between training and test data, maximum a posteriori classifiers were constructed based on Gaussian models arrived at through Fisher multiple‐discriminant analysis. The classifiers distinguished transient from continuant tones with approximately 99% correct performance. Instrument families were identified with approximately 90% performance, and individual instruments were identified with an overall success rate of approximately 70%. These preliminary analyses compare favorably with human performance on the same task and demonstrate the utility of the hierarchical approach to classification.

Automatic transcription of simple polyphonic music

A novel computational system has been constructed which is capable of transcribing piano performances of four‐voice Bach chorales written in the style of 18th century counterpoint. The system is based on the ‘‘blackboard’’ architecture, which combines top‐down and bottom‐up processing with a representation that is natural for the stated musical domain. Knowledge from auditory physiology, physical sound production, and musical practice has been successfully integrated in the current implementation, enabling the system to infer the presence of octaves, thereby overcoming a common pitfall of many traditional bottom‐up transcription systems. The architecture and incorporated knowledge will be presented, along with an appraisal of the system’s performance and comparisons with other systems in the literature. [Work supported by NSF.]

A computational model of auditory localization

A novel computational model of localization has been implemented. Using a simplified model of the cochlea filter bank, maximum‐likelihood estimates of azimuth and elevation are formed from interaural intensity and phase differences at energy onsets in multiple critical bands. Preliminary results show that the model is capable of resolving front/back confusions and other elevation ambiguities resulting from the ‘‘cones of confusion.’’ The onset sampling strategy is intended to model psychoacoustic phenomena such as the precedence effect and allows the model to deal with multiple sound sources and reverberant environments. Examples of the model’s ability to resolve ambiguous signals will be presented and compared with human performance. [Work supported by NSF.]