Signal anomaly detection using modified CUSUM method

An important aspect of detection of anomalies in signals is the identification of changes in signal behavior caused by noise, jumps, changes in bandwidth, sudden pulses and signal bias. A methodology is developed to identify, isolate and characterize these anomalies using a modification of the cumulative sum (CUSUM) approach. The algorithm performs anomaly detection at three levels and is implemented on a general-purpose computer