A Novel Model Research Of Feature Extraction For Weak EEG Signals

Electroencephalography (EEG) is the electrical activity of scalp on the surface caused by billions of cerebral neurons, which contains an abundant of physical, psychological and pathological information. Therefore, the acquisition, analysis and processing of EEG have a significance either in treatment and diagnosis of brain diseases, or in research of brain cognitive science. At present, reported by national and international documents, the feature extraction of single-channel EEG is not obvious, which cannot reflect features with non-stationary and chaotic of brain information effectively. It makes the application range of hidden features of EEG limited.In this paper, the integrated weak EEG analysis system was studied and developed with a new perspective of analysis model, which was completely different from the way of traditional EEG analysis. A variety of functions had been implemented by the system. First, the acquisition system with specific parameters was developed, which could collect richer depth of hidden weak EEG. The system consisted of various analysis methods, such as time domain, frequency domain, time-frequency, nonlinear, spectral analysis, chaotic operator, other statistical processing algorithms and so on, which can be selected to construct the integrated analysis model. It is effective of extracting the features for single-channel weak EEG which extended the depth of EEG analysis. And the asynchronous comparison analysis of dual-channel was introduced for comparing the characteristics between different signals or the different stages of the same signal by the feature extraction model of EEG for different applications. In the functional verification experiments, the functional verification of features recognition was achieved by the integrated analysis and the dual-channel analysis for epilepsy, sleep and depression EEG signals which had been confirmed by the special pre-study in our labs. The characteristics between different frequency bands of EEG were also distinguished. It is illustrated that the new analysis system developed can acquire the features of EEG in the state of time domain, frequency domain, nonlinear, chaotic, and spatial synchronously, which achieve the feature recognition of single-channel EEG and the parallel research of multivariate parameters. It also provides the basis for clinical disease diagnosis and research of brain science. The system is an open framework, it will be continue to develop the new analysis methods of the weak signal, and integrated more individuation analysis model aiming at different applications, as well as syncretize more signal processing methods in the acquisition system which could achieve the real-time analysis and dynamic monitoring of EEG. It is no doubt that this novel analysis system has a wide application prospect in the field of disease diagnosis and monitoring, brain science, brain cognitive research and so on.