AUDITORY INFORMATION PROCESSING USING SPIKING NEURAL NETWORKS

Abstract

Throughout the history of machine audition research, neuroscience has played a pivotal role by offering a bountiful source of inspiration for novel acoustic features, algorithms, and architectures. To achieve more capable, efficient, and reliable machine auditory systems, it calls for continuously exchange ideas between the fields of neuroscience and AI. The brain-inspired spiking neural network (SNN), which is considered as the third generation of neural network models, has shown great potentials for high performance, energy-efficient computing. In this thesis, we develop auditory neural codes and spike-based learning algorithms that are fundamental to efficient and effective pattern recognition using SNNs. Grounded on these studies, we develop SNN-based auditory systems for accurate, rapid, and efficient auditory information processing, including automatic speech recognition, environmental sound classification, and speech separation.