Speech Segmentation and Clustering Methods for a New Speech Recognition Architecture

To reduce the gap between performance of traditional speech recognition systems and human speech recognition skills, a new architecture is required. A system that is capable of incremental learning offers one such solution to this problem. This thesis introduces a bottom-up approach for such a speech processing system, consisting of a novel blind speech segmentation algorithm, a segmental feature extraction methodology, and data classification by incremental clustering. All methods were evaluated by extensive experiments with a broad range of test material and the evaluation methodology was itself also scrutinized. The segmentation algorithm achieved above standard quality results compared to what is found in current literature regarding blind segmentation. Possibilities for follow-up research of memory structures and intelligent top-down feedback in speech processing are also outlined.