Mining Speech: A Computational Model of Human Perception

The paper at hand deals with the complexity and variability of speech signals and presents the multilateral computational model of speech processing. We propose a speech mining approach of analyzing multidimensional audio data through identifying the nature of information carried out in speech, mechanisms of its extraction, and finally, mechanisms of discovering patterns, dependencies, relatedness, and trends. The integration of different approaches from speech processing, acoustics, phonetics, natural language processing, neurophysiology, and data mining provides the technology for comprehensive analysis of speech that can be applied in a variety of research domains as well as in business and government organizations.