Presentation | 2020-01-10 Examination of abnormal sound detection method of machine using standard deviation of multiple models Akihio Ito, HIroyuki Nishi, Manbu Okamoto, |
---|---|
PDF Download Page | PDF download Page Link |
Abstract(in Japanese) | (See Japanese page) |
Abstract(in English) | Along with the progress of super aged society, the number of elderlies living alone has increased. Therefore, the demand for services to watch the daily life of the elderly is increasing. Instead, video cameras and pendant sensors have problems such as privacy infringement and annoyance to wear things. Therefore, the authors have proposed a watching system that detects abnormal life sounds using a neural network. However, it is extremely difficult to collect all the abnormal life sounds for learning. Therefore, if the input sound has been learned in advance or not, and if an unlearned sound (unlearned sound) is detected, the user is designated as normal and normal. By adding the sound to the learning sound, the learning sound is expanded, and if there is no input that it is normal, a system that judges that it is abnormal is examined. Ten models with different initial weight values were prepared, and the experiment showed that the standard deviation of the output values differed between learned and unlearned. As a result of experiments, the performance improved compared to the conventional method. In this study, we use the public data set ToyADMOS to detect abnormal sounds and unlearned sounds. |
Keyword(in Japanese) | (See Japanese page) |
Keyword(in English) | Elderly people / Observation / Life sound / Neural network / Untrained acoustic data |
Paper # | ICM2019-37,LOIS2019-52 |
Date of Issue | 2020-01-02 (ICM, LOIS) |
Conference Information | |
Committee | LOIS / ICM |
---|---|
Conference Date | 2020/1/9(2days) |
Place (in Japanese) | (See Japanese page) |
Place (in English) | ARKAS SASEBO |
Topics (in Japanese) | (See Japanese page) |
Topics (in English) | Practical Use of Lifelog, Office Information System, Business Management, etc. |
Chair | Tomohiro Yamada(NEL) / Kiyohito Yoshihara(KDDI Research) |
Vice Chair | Toru Kobayashi(Nagasaki Univ.) / Takumi Miyoshi(Shibaura Inst. of Tech.) / Yoichi Sato(Open Systems Laboratory) |
Secretary | Toru Kobayashi(Research Organization of Information and Systems) / Takumi Miyoshi(NTT) / Yoichi Sato(NTT) |
Assistant | Kenichi Arai(Nagasaki Univ.) / Hiroki Nakayama(Bosco) |
Paper Information | |
Registration To | Technical Committee on Life Intelligence and Office Information Systems / Technical Committee on Information and Communication Management |
---|---|
Language | JPN |
Title (in Japanese) | (See Japanese page) |
Sub Title (in Japanese) | (See Japanese page) |
Title (in English) | Examination of abnormal sound detection method of machine using standard deviation of multiple models |
Sub Title (in English) | |
Keyword(1) | Elderly people |
Keyword(2) | Observation |
Keyword(3) | Life sound |
Keyword(4) | Neural network |
Keyword(5) | Untrained acoustic data |
1st Author's Name | Akihio Ito |
1st Author's Affiliation | Sojo University(Sojo Univ..) |
2nd Author's Name | HIroyuki Nishi |
2nd Author's Affiliation | Sojo University(Sojo Univ..) |
3rd Author's Name | Manbu Okamoto |
3rd Author's Affiliation | Sojo University(Sojo Univ..) |
Date | 2020-01-10 |
Paper # | ICM2019-37,LOIS2019-52 |
Volume (vol) | vol.119 |
Number (no) | ICM-358,LOIS-359 |
Page | pp.pp.39-44(ICM), pp.39-44(LOIS), |
#Pages | 6 |
Date of Issue | 2020-01-02 (ICM, LOIS) |