Conference proceeding
A study of a cross-language perception based on cortical analysis using biomimetic STRFs
INTERSPEECH 2019, v 2019-, pp 1971-1975
01 Jan 2019
Abstract
For those in the early stage of learning a foreign language, they commonly experience difficulties in understanding spoken words in the second language, while they have no problem in recognizing words spoken in their mother tongue. This paper examines this phenomenon using biomimetic receptive fields that can be interpreted as a transfer function between acoustic stimulus and cortical responses in the brain. While receptive fields of individual subjects are often optimized to recognize unique phonemes in their mother language, it is unclear whether challenges associated with acquiring a new language (especially in adulthood) is due to a mismatch between phonemic characteristics in the new language and optimized processing in the system. We explore this question by contrasting biomimetic systems optimized for four different languages with sufficiently different characteristics. We perform English phoneme classification with these language-optimized systems. We observed distinctive characteristics in receptive fields emerging from each language, and the differences of English phoneme recognition performance accordingly.
Metrics
3 Record Views
Details
- Title
- A study of a cross-language perception based on cortical analysis using biomimetic STRFs
- Creators
- Sangwook Park - Johns Hopkins UniversityDavid K. Han - Army Res Lab, Computa & Informat Sci Directorate, Aberdeen Proving Ground, MD USAMounya Elhilali - Johns Hopkins UniversityInt Speech Commun Assoc
- Publication Details
- INTERSPEECH 2019, v 2019-, pp 1971-1975
- Series
- Interspeech
- Publisher
- Isca-Int Speech Communication Assoc
- Number of pages
- 5
- Grant note
- N000141612879; 629 N000141912014; N000141712736 / Office of Naval Research U01AG058532; R01HL133043 / National Institutes of Health; United States Department of Health & Human Services; National Institutes of Health (NIH) - USA
- Resource Type
- Conference proceeding
- Language
- English
- Academic Unit
- Electrical and Computer Engineering
- Web of Science ID
- WOS:000831796402022
- Scopus ID
- 2-s2.0-85074712614
- Other Identifier
- 991021931084104721
InCites Highlights
Data related to this publication, from InCites Benchmarking & Analytics tool:
- Collaboration types
- Domestic collaboration
- Web of Science research areas
- Audiology & Speech-language Pathology
- Computer Science, Artificial Intelligence