Conventional methods for source localization requires large number of sensors for resolving closely spaced sources. Also, it gives many spurious Direction of Arrival (DOA) peaks which makes decision of DOA candidate peak difficult. Under such circumstances, tracking or Distant Speech Recognition of sources becomes difficult. Achieving high resolution source localization and tracking in spherical harmonics domain is objective of this proposal. In particular, the aim of the proposed research is to study the MUSIC-Group delay (MUSIC-GD) spectrum in spherical harmonics domain. MUSIC-GD spectrum is a novel, high resolution subspace-based method for source localization. This Proposed research will try to utilize the advantages of the MUSIC-GD spectrum in a spherical harmonics framework, being computationally simple and more accurate.
The objectives of the project are:

1. Localization of speech sources
2. Three dimensional real time tracking of speech sources
3. Beamforming and Speech enhancement
4. Distant speech recognition (DSR)
5. Room acoustics measurement
6. 3-D positional audio and head related transfer function (HRTF) measurement for rendering spatial audio.

Highlight of the Project

This will facilitate and augment a dual spherical microphone array set-up at IIT Kanpur probably the first in India. This will enable the research initiative on spherical array processing.