Monaural localization using combination of TD (λ) with backpropagation

The problem of monaural localization is centered around that of determining the sharpest outcome of the various allowable deconvolutions. Short-time processing is used to extract the sharpness features from the deconvolutions and produces new sequences of feature pattern. The method of combination of TD (λ) with backpropagation is used for this sequential pattern recognition. Preliminary experimental results show that the method is capable of recognizing the sharpest outcome, and therefore provides the estimation of sound source direction. The method is potentially applicable to the case where pattern recognition is made based on a series of observations or a sequence of feature patterns