A high-efficiency RISC-V microprocessor architecture optimized for audio noise suppression

Audio noise suppression techniques are commonly utilized across diverse audio systems,often relying on general-purpose processors. However, this leads to inefficiencies in terms of hardware resource usage, energy consumption, and overall system expense. This paper presents a specialized microcontroller for audio noise suppression,based on the Hummingbird E203 framework. Key hardware optimizations include enhanced multiplication and division units using Booth4 encoding, Wallace tree, and dual SRT-4 mechanisms, alongside the integration of an F instruction set for floating-point operations. A camouflage strategy reduces memory transaction complexity. Software improvements utilize adaptive spectral subtraction for efficient noise suppression. Benchmarks show a 4002.5% boost in floating-point performance and significant gains in noise reduction, with a 54.6% decrease in noise energy and a 28.7% rise in signal-to-noise ratio, highlighting its efficacy.