A Novel Method of Absolute Digital Demodulation

This paper proposes a digital demodulation method with all functions realized in digital processing algorithms. The bandpass digital modulation signals are transformed to direct digital sequence in the first step. After adaptive filtering, a novel bit period infimum calculation algorithm is used to classify the modulation patterns. According to the relationship between the infimums of the direct digital sequence and it´s recognition sequence, different modulation patterns can be classified easily. The probability of the correct recognition of the algorithm is higher than 95% at low signal-noise ratio condition. A 3-layer backpropagation neural network based demodulation algorithm is proposed to transform the direct digital sequence to digital baseband signal. For different patterns of the digital modulation signal, different artificial BP neural networks are designed to achieve good demodulation performance