Room localization for distant speech recognition

Result of the Month

BDiagr.png

The problem of room localization is to determine where, in a multi-room environment, a person is producing a speech utterance. At Interspeech 2014 we have presented the system of the figure. It exploits the information gained from a network of microphones installed in house, where the lack of calibration of the microphone energies creates an additional challenge.

Contact: Juan Andrés Morales Cordovilla


The Word-Accuraccy (WAcc) of the baseline (based on just identifying the room where the VADs detects the maximum energy) is 79 %. The WAcc of the proposed system (based on a LDA classifier with high-SNR-energy+coherence as input feature) improves to 90%.

1. November 2014 - 30. November 2014