Demo Abstract: Acoustic Signal Processing For Anomaly Detection In Machine Room Environments
BuildSys '16: The 3rd ACM International Conference on Systems for Energy-Efficient Built Environments Palo Alto CA USA November, 2016(2016)
摘要
We present a system that uses acoustic signals to monitor equipment in commercial buildings, such as in machine rooms with HVAC system components. The system uses an ensemble of machine learning classifiers to effectively label signals as either "normal" or "abnormal". We collect audio clips from mobile devices in a machine room and an elevator shaft in the main building of IBM Research. We use these to learn the spectrum of normal sound signatures and identify abnormal sounds that fall outside this range. Abnormal sounds detected by the system are presented to the end user for anomaly confirmation. We also integrate a work order system to automatically issue a repair work-order if the sound is abnormal.
更多查看译文
AI 理解论文
溯源树
样例
生成溯源树,研究论文发展脉络