Hiroshi Sawada

Nara Institute of Science and Technology, NTT (Japan)

Papers

Total Citations

H-Index

About

Hiroshi Sawada is a leading researcher in robot auditory systems, specializing in speech enhancement and robust human-robot interaction. His work focuses on solving the critical challenge of enabling robots to understand human speech amidst their own internal mechanical noise. Sawada pioneered semi-blind signal processing techniques that leverage additional sensors inside robots to record and suppress internal noise, dramatically improving speech recognition in hands-free dialog systems. His 2009 paper on this architecture, which has garnered 14 citations, introduced a modified frequency-domain blind signal separation method that remains influential in the field. Recognizing the limitations of cascaded auditory functions, Sawada later proposed a groundbreaking unified framework based on Bayesian topic models, integrating sound source localization and separation into a single, robust system less prone to cascading errors. Further advancing practical applications, he developed end-fire microphone array configurations specifically designed to discriminate between target speech direction and robot noise, achieving notable improvements in real-world spoken dialog systems. Through these contributions, Sawada has established himself as a key innovator in creating more reliable, noise-resilient auditory perception for interactive robots.