Automatic speaker recognition is a complex and multifaceted problem that combines aspects of several disciplines of science and engineering. Over the last few decades, a wide variety of techniques have been employed in attempts to solve this problem.; A survey of the literature of speech and speaker recognition was made and is summarized. An isolated word, text-dependent recognizer consisting of a cepstral processing based front-end which extracts a minimal feature set coupled with a two-layer feedforward neural network for robust classification was designed and implemented. A set of recognition utterances was selected and a pool of ten volunteers was recruited and used for testing.; False acceptance rates of 2-6% were achieved for single word based networks. This compares well with results reported by others using similar techniques, and demonstrates the potential of this approach. |