Multithreaded Convolution Implementation Based on Block Methods

A multithreaded convolution implementation based on block algorithms is considered. Convolution essentially the basis of many methods that solve problem determining degree similarity or independence two processes, in other words, when it necessary to determine correlation. The algorithm itself executed with a significant delay, because for its execution accumulate entire signal and then process it. analysis showed one possible ways reduce time costs algorithms. article shows main features by overlap method addition addition, as well numerical examples. results obtained show application these without use window function leads distortions spectrum. Based analysis, universal scheme performing processing an input data proposed. This allows achieve good compromise between computational complexity, system architecture, costs.