Feature extraction is the most important step in effective state monitoring of analog circuit.Due to variety of features,the paper proposes a new approach to monitor state of analog circuit: Firstly,composite features including time-domain features and wavelet features are extracted from the response signal through giving different stimulus to analog circuit; Secondly,HMMs trained by composite features are utilized to recognize different faults,thus we can acquire effective state monitoring of analog circuit.The proposed approach is applied to a typical analog circuit and experimental results show that the proposed approach provides a better monitoring ability by comparing with BP network and offers a practical method for state monitoring of analog circuit.
展开▼