ABSTRACT
We present the major results of our research in the area of speech recognition in the presence of sporadic environmental sounds. We refer to this general problem area as environmental speech recognition (ESR). Our ESR-related research, which took the form of several case studies, has led to the development of a unified terminology that can be used to describe acoustic scenarios which contain both speech and environmental sounds. We have also used the case studies to examine the viability of a novel knowledge-based approach to ESR problems. The framework for Integrated Processing and Understanding of Signals (IPUS) has been identified as a suitable system architecture for implementing the knowledge-based approach.
