Systems for robust speech recognition in far-field scenarios
We are working on automatic transcription of natural conversations, which can range from informal get-togethers to professional meetings. We focus on either using single-channel or multi-channel recordings to achieve a robust preprocessing and speech enhancement that allows for accurate transcription of current speech recognition systems even in acoustically challenging environments. If you are interested in advancing methods, unorthodox applications of neural networks and would like to apply your math and programming knowledge in these areas, please visit us in our office so that we may discuss possible topics.
