Characteristic recognition of pulse in frequency domain based on probabilistic neural network

In order to improve the level of automation for pulse signal processing and recognition, firstly, it used the fast Fourier transform with radix-2FFT algorithm to extract the characteristic parameters of pulse in frequency domain. Secondly, it made the dyadic discrete wavelet transform in four scales with MALLAT algorithm and calculated the detail energy values of four scales to constitute together the feature vector of pulse. Finally, for the shortcomings of traditional identification methods, the probabilistic neural network pulse identification method was proposed. It designed probabilistic neural network classifier and made the classification experiment, and the recognition rate was 93.00%. The results showed that the extracted feature vectors have a strong description capability of pulse.