Jump resonance is a phenomenon observed in nonlinear circuits where the output exhibits abrupt jumps when the frequency of the input signal is varied. In literature, several methods have been proposed for modeling and predicting of jump-resonance, which has led to circuit designs that are optimized to avoid this non-linear phenomenon. In this paper we propose exploiting jump-resonance based hystresis, observed in silicon cochlea, for encoding frequency and specifically formant trajectories in speech signal. Using experimental prototypes fabricated in a 0.5µm CMOS process, we show that the features extracted from a jump-resonance based silicon cochlea are more discriminative for speech based biometrics as compared to features extracted from a conventional silicon cochlea.
展开▼
