Auditory-inspired adaptive frequency tracking and harmonic grouping

One of the best sound analyzers to date is the human (mammalian) auditory system, which has evolved through millions of years by resolving classification problems. It is a versatile, elegant. and powerful sound processing unit. It excels in detecting, estimating, and classifying multiple targets simultaneously even in noisy environments. Hence, mimicking even some known features of the auditory system could be beneficial in developing improved frequency tracking and classification algorithms. An auditory inspired adaptive synchrony capture filterbank (SCFB) signal processing architecture for tracking signal frequency components was proposed in an earlier paper [JASA-2013]. The SCFB exhibits many desirable properties for processing speech, music, and other complex sounds. The algorithm was modified using adaptive tuning parameters, and a generalized way to determine and suppress voiced and unvoiced (silent) regions. This modified algorithm estimates frequencies with higher accuracy even in the presence of closely spaced input tones [ASA-2024 presentation]. Recent work extended mimicking the auditory system further and in the process made the algorithm computationally efficient. In addition, the estimated frequency tracks are harmonically grouped (or clustered), which is a first step in sound source (or signal) separation. Preliminary analysis shows promising results. This presentation will focus on these latest updates.