2017 Fiscal Year Final Research Report
Study on Detection of Abnormality in Daily Life by Frequency Analysis of Indoor Activity Sound
Project/Area Number |
15K00746
|
Research Category |
Grant-in-Aid for Scientific Research (C)
|
Allocation Type | Multi-year Fund |
Section | 一般 |
Research Field |
Clothing life/Dwelling life
|
Research Institution | Akita University |
Principal Investigator |
Tanaka Motoshi 秋田大学, 理工学研究科, 准教授 (50261649)
|
Project Period (FY) |
2015-04-01 – 2018-03-31
|
Keywords | 生活活動音 / 信号解析 / 異常検出 / 確率モデル |
Outline of Final Research Achievements |
In order to investigate on a detection method of abnormality in daily life for persons living alone, time-frequency analysis of indoor activity sound recorded with 196 kHz sampling frequency and 80 kHz bandwidth was demonstrated. The frequency band of power spectrum up to 80 kHz was divided into fifteen subbands with bandwidth of 10 kHz, and the power in each subband, as a feature vector, was calculated. Using these feature vectors, stochastic modeling of daily life sound was discussed, and generation probability of the sound was obtained with the stochastic model. Then the detection algorithm of the abnormality was discussed using spectral flux, Mahalanobis’ distance of 12 MFCCs (Mel Frequency Cepstral Coefficients), and generation probability of the sound, and fifteen sounds which included simulated abnormal situations were tested. The result indicated the feasibility of the detection method combining those three parameters.
|
Free Research Field |
情報通信工学
|