Speech and Microphone Array Research

 

Overview:

The Speech and Signal Processing activity in LEMS is essentially the research area of Professor Harvey Silverman, with some substantial collaborations with William Patterson. Professor Silverman has graduated 27 PhD’s in his 32-year tenure at Brown and currently has a group of three PhD students. After much work in the areas of speech recognition and hardware for speech processing, reconfigurable and parallel computing and microphone array research, the current activity is focused on addressing many of the problems associated with real environments using large-aperture microphone arrays. There is also work in the embedded signal-processing systems area. Professor Silverman has been a strong supporter of undergraduate research experiences and has been the mentor for a large number of independent studies projects and Honors’ these over the years.

Research

Other research areas include speech recognition, microphone-array signal processing, reconfigurable computing, etc. Our research focuses on designs and algorithms for real-time digital signal processing/speech processing-based systems. In our lab, a huge microphone-array (HMA) system that supports real-time processing and data acquisition for 512 microphones has been in operation for more than ten years. It is still in use for algorithm research and data acquisition. A new microphone array (HMAII) of 128 microphones has been in operation for almost five years. We implemented a new locationing algorithm (SRP-PHAT using SRC) in real-time on this new array. A new method for automatic array calibration called the Free-Source Method (FrSM) was recently used to precisely obtain the three-dimensional cooridnates of the microphones in the array. In the near future, we would like to design a teleconfrencing environment between the two arrays.

    Fish-eye view of HMAIIFish-eye view of HMAII

We have also successfully built the first version of our wireless microphone array that can communicate with each other and the recording PC without the aid of lengthy cables. The wireless microphone array is composed of multiple self-contained units that have both a microphone and a speaker. We recently implemented a self-calibration algorithm using multidimensional scaling (MDS) on the wireless micorphone array. In the future, we hope to reduce the size of the units and mitigate other issues with our current version. Such a wireless system is very flexible in deployment, easily scalable, easy to maintain, and quick to set up.

Both the wired and the wireless microphone arrays have wide applications in the fied of teleconferencing, speech recognition, speaker identification, acoustic surveillance, sound capture in reverberant conditions etc.

Related Publications