On the Role of Subglottal Acoustics in Height Estimation, and Speech and Speaker Recognition
May 14, 2014
from 12:00 PM to 02:00 PM
|Where||Engr. IV Bld., Faraday Room 67-124|
|Contact Name||Harish Arsikere|
|Add event to calendar||
Adviser: Prof. Abeer Alwan
The subglottal system comprises the trachea, bronchi and their accompanying airways. Its configuration changes very little compared to that of the supraglottal vocal tract, as a result of which its acoustic properties are relatively more stationary and speaker specific. In this work, our knowledge of subglottal acoustics---subglottal resonances (SGRs), most importantly---is leveraged to develop novel solutions to three problems, all of which involve using or estimating speaker-specific characteristics: (1) height estimation, (2) speaker normalization for automatic speech recognition (ASR), and (3) speaker identification (SID) and verification (SV). The focus is on scenarios where purely statistical methods may be sub-optimal owing to limited and/or noisy speech data.
This talk covers the following topics: (1) collection of new databases comprising simultaneous recordings of speech (microphone) and subglottal acoustics (accelerometer); (2) data analysis to understand the relationships among SGRs, vocal-tract resonances, body height and language; (3) automatic algorithms for estimating SGRs from speech signals (i.e., without using accelerometer information); and (4) knowledge-based methods and empirical results---for height estimation, ASR, SID and SV tasks---that demonstrate the efficacy of SGRs and other subglottal features in noisy environments and limited-data scenarios. In general, the proposed methods are significantly better than conventional approaches in terms of performance, data requirements and generalizability.
This work was supported in part by the National Science Foundation.
Harish Arsikere is a PhD candidate in the Electrical Engineering Department at UCLA. Prior to this, he obtained a Master of Technology degree from Indian Institute of Technology, Kanpur, in 2009, and a Bachelor of Engineering degree from R. V. College of Engineering, Bangalore, in 2007. He received the UCLA EE Department Fellowship for top scoring in the 2011 PhD Prelim Exam. Harish is broadly interested in signal and speech processing with a particular liking for knowledge-based solutions.