A new approach based on joint entropy maximization (JEM) is taken and adaptive algorithms are developed for channel equalization with a decision feedback equalizer (DFE). The proposed work generalizes the existing algorithms for DFE with a hard decision device. Previous research has shown that when the hard decisions in a DFE are replaced with soft decisions, the performance of the adaptive algorithms [e.g., minimum mean square error (MMSE)] improves dramatically. The soft decisions can be introduced naturally via the viewpoint taken here. Additionally, constant modulus and other (blind) algorithms for DFE with soft decisions can be derived from this JEM approach.

Along with the spoken message, speech contains information about the identity of the speaker. Thus, the goal of speaker identification is to develop features which unique to each speaker. This paper explores a new feature for speech and shows how it can be used for robust speaker identification. The results are compared to the cepstrum feature due to its widespread use and success in speaker identification applications. The cepstrum, however, has shown a lack of robustness in varying conditions, especially in a cross-condition environment where the classifier has been trained with clean data but then tested on corrupted data. Part of the bispectrum is used as a new feature and we demonstrate its usefulness in varying noise settings.

The separation of multiple signals from their superposition recorded at several sensors is addressed. The methods employ polyspectra of the sensor data in order to extract the unknown signals and estimate the finite impulse response (FIR) coupling systems via a linear equation based algorithm. The procedure is useful for multichannel blind deconvolution of colored input signals with (possibly) overlapping spectra. An extension of the main algorithm, which can be applied for quasiperiodic signal separation, is also given. Simulation results corroborate the applicability of the algorithm.

Fourth-order statistics based methods are proposed for estimating the range and bearing of multiple near-field sources. The performance of the maximum likelihood estimator based on the asymptotic distribution of the sample trispectrum is studied and it is shown that the proposed estimator has lower variance when compared to the Cramer-Rao bound. This suggests that under high SNR, and with nonGaussian signals, the nonlinear HOS based method provides better performance than techniques that employ either correlations or spectra. Simulations are provided to illustrate the theoretical results.