Higher Order Crossings Analysis Of Signals Over Graphs

We propose the extension of the Higher Order Crossings (HOC) analysis concept for signals over graphs. The HOC over graphs sequence, (HOCg), is defined and its efficacy on graph signal discrimination is investigated. Compared with state of the art graph signal discrimination algorithms, the proposed HOCg combined with a Support Vector Machine classifier achieves fast and similar or better performance over five different benchmark data sets.