Feature Extraction from Digital Communications Signals Using Wavelet Transforms

摘要

Research was conducted to evaluate the feasibility of applying wavelet transformsand methods based on proportional bandwidth processing to transient feature extraction. Classical wavelets and signal related basis functions were used to extract the switching instants from BPSK, QPSK, FSK, ASK, and AMQPSK signals in the presence of additive Gaussian white noise. Results show that using the magnitude of one output scale, the transient times can be detected above a certain signal to noise ratio level. This level depends on the type of signal and wavelet, as well as on the acceptable number of mistakes. Additional work, using several scales and a properly trained neural network, should demonstrate automated selection of the modulation type.