Unsupervised Speech Controlled System with Long-Term Adaptation

Abstract

Chapters 2 and 3 provided a general overview of problems related to the scope of this book. The fundamentals and existing solutions as known from literature were discussed in detail.

In the preceding chapter a combination of short-term and long-term adaptation was examined. Speaker profiles can be initialized and continuously adapted under the constraint that the speaker has to be known a priori. The experiments carried out showed a significant improvement for speech recognition if speaker identity is known or can be at least reliably estimated.