Study examines feasibility of DNN-based speech enhancement on FPGA for hearing aids

Arxiv CS.SD (Sound/Audio ML)

Paper characterizes the latency and power gap between DNN-based speech enhancement and hearing aid constraints. Lightweight models for speech separation and denoising are deployed on embedded FPGA.

speech-enhancement

Arxiv CS.SD (Sound/Audio ML)·Feyisayo OlalereUmut AltinKiki van der HeijdenMarcel van Gerven

7 days ago