Waveform recognition method based on deep neural network under small sample condition

The invention discloses a waveform recognition method based on a deep neural network under a small sample condition. The waveform recognition method comprises two parts of small sample amplification and waveform recognition. The small sample amplification comprises the steps of determining the minimum sampling sample number required by the amplification signal and realizing the small sample amplification by utilizing a probability density function estimation method. The invention comprises the following steps: firstly, determining a minimum sampling sample number required by a signal to be amplified through a KLIEP algorithm; then estimating a probability density function of a signal to be amplified through the minimum sampling sample, and completing amplification of the small sample according to the estimated probability density function; and finally, realizing waveform identification by using the amplified signal and the deep neural network. According to the method, the waveform recognition performance unde