Abstract
This paper presents an adaptive maximum entropy quantizer for fault detection in smart sensors. An unsupervised learning rule adapts the quantization thresholds for obtaining equal-probability quantization intervals, therefore giving a reliable non-parametric density estimation. Hypothesis testing based on the likelihood ratio is then applied for detecting abrupt changes in the signal distribution which reflects faulty sensor operations.
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have