1 paper
Sri Harsha Dumpala, Dushyant Sharma, Chandramouli Shama Sastri +3
We present a novel method for extracting neural embeddings that model the background acoustics of a speech signal. The extracted embeddings are used to estimate specific parameters…