Presentation | 2016-08-09 Improvement to an Objective Binaural Intelligibility Prediction Method Kazuya Taira, Kazuhiro Kondo, |
---|---|
PDF Download Page | PDF download Page Link |
Abstract(in Japanese) | (See Japanese page) |
Abstract(in English) | We attempted to improve the binaural intelligibility estimation method proposed in our previous paper. The number of data in the lower subjective intelligibility range which had been missing was supplemented performing additional evaluation for a more uniformly distributed data. Also, we tried four feature value extraction methods, such as the Band-Selection Better Ear Model. Furthermore, we tried estimation using machine learning in addition to regression analysis which we applied previously. Neural network, support vector machine and random forest were used here. Estimation using random forest showed significantly higher estimation accuracy compared to other previous results, with RMSE of 0.1 or less, and correlation of 0.9 or more in closed tests. Estimation using random forest also yielded a higher accuracy than that of logistic regression analysis in open test, in which the accuracy was shown to decrease significantly in previous tests. |
Keyword(in Japanese) | (See Japanese page) |
Keyword(in English) | Speech Intelligibility / Intelligibility Estimation / Binaural Hearing / Japanese Diagnostic Rhyme Test / Logistic Regression Analysis / Machine Learning |
Paper # | EA2016-21 |
Date of Issue | 2016-08-02 (EA) |
Conference Information | |
Committee | EA / ASJ-H |
---|---|
Conference Date | 2016/8/9(2days) |
Place (in Japanese) | (See Japanese page) |
Place (in English) | Tohoku Gakuin Univ., Tagajo Campus |
Topics (in Japanese) | (See Japanese page) |
Topics (in English) | Ultra-Realistic Acoustics, Engineering/Electro Acoustics, Psychological and Physiological Acoustics, and Related Topics |
Chair | Mitsunori Mizumachi(Kyushu Inst. of Tech.) / Tatsuya Hirahara(富山県立大) |
Vice Chair | Yoichi Haneda(Univ. of Electro-Comm.) / Suehiro Shimauchi(NTT) / Shigeto Furukawa(NTT) |
Secretary | Yoichi Haneda(KDDI R&D Labs.) / Suehiro Shimauchi(Akita Pref. Univ.) / Shigeto Furukawa(リオン) |
Assistant | Shigeto Takeoka(Shizuoka Inst. of Science and Tech.) / TREVINO Jorge(Tohoku Univ.) |
Paper Information | |
Registration To | Technical Committee on Engineering Acoustics / Auditory Research Meeting |
---|---|
Language | JPN |
Title (in Japanese) | (See Japanese page) |
Sub Title (in Japanese) | (See Japanese page) |
Title (in English) | Improvement to an Objective Binaural Intelligibility Prediction Method |
Sub Title (in English) | |
Keyword(1) | Speech Intelligibility |
Keyword(2) | Intelligibility Estimation |
Keyword(3) | Binaural Hearing |
Keyword(4) | Japanese Diagnostic Rhyme Test |
Keyword(5) | Logistic Regression Analysis |
Keyword(6) | Machine Learning |
1st Author's Name | Kazuya Taira |
1st Author's Affiliation | Yamagata University(Yamagata Univ.) |
2nd Author's Name | Kazuhiro Kondo |
2nd Author's Affiliation | Yamagata University(Yamagata Univ.) |
Date | 2016-08-09 |
Paper # | EA2016-21 |
Volume (vol) | vol.116 |
Number (no) | EA-179 |
Page | pp.pp.7-12(EA), |
#Pages | 6 |
Date of Issue | 2016-08-02 (EA) |