The Cutting Edge of Speech Recognition

In this episode of the IEEE Signal Processing Society podcast, Dr. Sanjeev Khudanpur, Director of the Center for Language and Speech Processing, Johns Hopkins University interviews Associate Prof. Shinji Watanabe, Language Technologies Institute, Carnegie Mellon University. They talk about the latest research and innovations in speech recognition technologies and their impact across various industries. Shinji WatanabeShinji Watanabe is an Associate Professor at Carnegie Mellon University in Pittsburgh and a leading researcher in speech and language processing. His work spans automatic speech recognition, speech enhancement, spoken language understanding, and machine learning for speech and language processing. He has contributed more than 500 publications to peer-reviewed journals and received several awards, including the best paper award from ISCA Interspeech 2024. In this episode, Associate Prof. Watanabe reflects on the transformative progress in speech recognition over the past decade, highlighting milestones from the adoption of deep neural networks to the rise of large-scale models like OpenAI Whisper. He discusses the ongoing challenges in achieving human-level understanding in complex scenarios such as multi-speaker conversations, accented and multilingual speech, and child or disordered speech. He concludes with thoughts on academia’s enduring role in shaping the field, and how his inspiration is often drawn from science fiction and Japanese animation.