A neural network approach for feature extraction and discrimination between Parkinsonian tremor and essential tremor

摘要: BACKGROUND: Essential tremor (ET) and the in Parkinson's disease (PD) are two most common pathological with a certain overlap clinical presentation.OBJECTIVE: The main purpose of this work is to use an artificial neural network select best features discriminate between types tremors using spectral analysis time-series recorded by accelerometry surface EMG signals.METHODS: Soft-Decision wavelet-based technique be used order obtain 16 bands approximate representation both accelerometer signals sets data (training test). training set consists 21 ET subjects 19 PD while test 20 subjects. has been for diagnostic purposes Department Neurology University Kiel, Germany. A type feed forward back propagation find frequency associated different that yield better discrimination efficiency on data. same designed set.RESULTS: Efficiency result 87.5% was obtained from each three under test.CONCLUSIONS: successfully feature extraction pattern matching tasks complete classification system.