Accession Number:

AD1033827

Title:

Language Recognition via Sparse Coding

Descriptive Note:

Technical Report

Corporate Author:

MIT Lincoln Laboratory Lexington United States

Report Date:

2016-09-08

Pagination or Media Count:

4.0

Abstract:

Spoken language recognition requires a series of signal processing steps and learning algorithms to model distinguishing characteristics of different languages. In this paper, we present a sparse discriminative feature learning framework for language recognition. We use sparse coding, an unsupervised method, to compute efficient representations for spectral features from a speech utterance while learning basis vectors for language models. Differentiated from existing approaches, we introduce a maximum a posteriori MAP adaptation scheme that further optimizes the discriminative quality of sparse-coded speech features. We empirically validate the effectiveness of our approach using the NIST LRE 2015 dataset.

Subject Categories:

  • Voice Communications

Distribution Statement:

APPROVED FOR PUBLIC RELEASE