Separation of Speech from Background
John R. Josephson, James Russell · 2004
Phase I has resulted in advances in computational methods, assessments of current strengths and weakness, and assessments of the range of potential applications and their performance requirements. The progress and results from Phase I show that the technology is very promising. However, it has also become clear that the problem is quite challenging, and significant technical weaknesses remain. Phase II would aim to overcome the most significant weaknesses preventing practical levels of performance. These are considered to be: inadequate performance on unvoiced speech materials, weak processing algorithms that mandate large amounts of computation and impose latencies that are too long for near-real-time applications.