Design and implementation of an efficient architecture for higher order statistics with DWT

Higher order statistical metric has the capability to extract the signal feature of non-stationary waveforms that the second order statistics can't. This paper proposes an efficient architecture for the frequency domain kurtosis metric, which is an example of the higher order statistics using DWT bit streams. The hardware design methodology emphasizes efficiency on area and time to meet the technological challenges of compactness and portability for biomedical applications.