Denoising in Low SNR Environment

An algorithm for clear conversation in noisy environments

Introduction


An algorithm that can make a conversation clearer in very noisy environments, such as construction sites and airports. In such environments, the sound is relatively weaker and typically drowned out by noise, which is known as low signal-to-noise ratio (SNR). This algorithm can help us better communicate in such environments. Details can be found in my PhD thesis.

Two workers on a chocolate construction site. How can they communicate easily in the presence of loud pile driver noise?

Demo


We simulated a scenario where a male and a female were speaking while their voices were drowned out by the sound of pile driver’s engine. After processing with our algorithm, we can clearly hear their talking. The algorithm used three microphones. The number of microphones is free to be reduced or increased (at least two).

noisy sound

processed sound