Sound Source Localization (SSL) is an important component to microphone array signal processing for speech applications. SSL serves as the front-end to acoustic beamforming by steering the beam to the desired sound source. The location estimation can also be used as a visual cue for smart and augmented reality devices.
The methods used for SSL vary from classical direction of arrival (DOA) solutions. Classical DOA approaches (Capon Minimum Variance, MUSIC) used in radar and sonar applications have the advantage of narrowband and statistically stationary far-field source signals with minimal multi-path reflections. Speech is a wideband non-stationary sound source, often located in reverberant environments.
To help combat the challenges of direction finding for speech applications, Sound Source Localization solutions consist of multiple stages. The first stage is to precondition the signal to focus on the frequencies of interest that have the best signal-to-noise ratio. The next step is to perform the time delay estimation and sound localization. The final stage is to cluster and smooth the multitude of estimates to make a final decision.
The engineering tradeoffs for SSL are quite straightforward. Adding more microphones into the system design adds redundancy and helps to improve accuracy. As one would expect, this will increase the computational complexity of the solution. VOCAL is an engineering design house, which can help guide you with the product design utilizing this technology.