Speech Analysis
Speech analysis is a rapidly evolving field focused on understanding and manipulating spoken language using computational methods, aiming to improve human-computer interaction and address challenges in healthcare and other domains. Current research emphasizes developing robust models, often based on transformer networks and neural codecs, for tasks such as speech recognition, emotion detection, and generation, including handling multi-speaker scenarios and low-resource languages. These advancements have significant implications for applications ranging from improved accessibility for individuals with speech impairments to more natural and intuitive interfaces for various technologies, as well as enabling new diagnostic tools in healthcare.
Papers
Echotune: A Modular Extractor Leveraging the Variable-Length Nature of Speech in ASR Tasks
Sizhou Chen, Songyang Gao, Sen Fang
Voxtlm: unified decoder-only models for consolidating speech recognition/synthesis and speech/text continuation tasks
Soumi Maiti, Yifan Peng, Shukjae Choi, Jee-weon Jung, Xuankai Chang, Shinji Watanabe