US 11,835,430 B2
Anomaly score estimation apparatus, anomaly score estimation method, and program
Yuma Koizumi, Tokyo (JP); and Shoichiro Saito, Tokyo (JP)
Assigned to NIPPON TELEGRAPH AND TELEPHONE CORPORATION, Tokyo (JP)
Appl. No. 17/630,112
Filed by NIPPON TELEGRAPH AND TELEPHONE CORPORATION, Tokyo (JP)
PCT Filed Jul. 30, 2019, PCT No. PCT/JP2019/029777
§ 371(c)(1), (2) Date Jan. 25, 2022,
PCT Pub. No. WO2021/019672, PCT Pub. Date Feb. 4, 2021.
Prior Publication US 2022/0283057 A1, Sep. 8, 2022
Int. Cl. G06F 11/30 (2006.01); G01M 99/00 (2011.01)
CPC G01M 99/005 (2013.01) 7 Claims
OG exemplary drawing
 
1. An anomaly score estimation apparatus which estimates an anomaly score of an observed signal, the anomaly score estimation apparatus comprising
processing circuitry configured to:
execute an anomaly score calculation processing, wherein the anomaly score calculation processing comprises correcting the anomaly score calculated from the observed signal, and wherein, when a similarity between the observed signal and a registered normal sound increases, a value of the anomaly score decreases; and
execute a feature extraction processing, wherein the feature extraction processing further comprises extracting a feature amount with a fixed-length from a time-series acoustic signal with an arbitrary length,
wherein the feature extraction processing is optimized to extract, from the time-series acoustic signal, the feature amount, the feature amount reflects a feature of an unknown normal sound, the time-series acoustic signal includes at least a known normal sound and the unknown normal sound,
the anomaly score calculation processing further comprises calculating a similarity score by a similarity score function, the similarity score function is learned by using the feature amount extracted from the time-series acoustic signal, and the time-series acoustic signal includes at least a normal sound.