As presented at MWSCAS/NEWCAS 2007, shown here is minor anomaly detection in an acoustic signal. Sucessive transformation of random variable resulted in a Teager energy outlier detection scheme in wavelet-filtered sub-bands. The red areas indicate where the total energy (potential & kinetic) deviated significantly when compared with the analysis window. The lower waveforms show four of eight wavelet decomposition levels with bandwidths and centering that reflect the dyadic sampling scheme.
huh?
psychedelos 4 years ago
The system detects unspecified anomalies. For example, if set up to monitor a room with people talking, and there is a sound that does not belong, such as a glass breaking, the system will detect it. It can also be used to monitor machines. This video, while public, was intended for a much smaller, private audience.
johnsalik 4 years ago