Current Work
The existing work in the literature is being reviewed. Source localization and beamforming was applied to existing TCS recordings. We initially take forward this work for a scenario of non-overlapping speakers. Post-filtering is done for single-channel recording of the microphone array using CNMF model with/without NMF model for speech. For illustration, sound files of a recorded non-overlapping speaker and it's beamformed output is given (obtained from the system mentioned in the proposed model):
| Distant speaker | |
| DSB (beamforming) | |
| CNMF-NMF (Post-filtering) |