ABR Signal Extraction Based On Wavelet And Improved Adaptive RBF Neural Network

The auditory brainstem response(ABR)refers to the electrical changes in the auditory pathway from the cochlea to the brainstem that the human ear can record between the top of the head and the mastoid process within 10 ms of receiving an auditory stimulus.The ABR signal consists of seven positive waves,named I through VII in the order in which they appear.Hearing and brain stem function can be diagnosed based on the presence or absence of each wave and the length of the incubation period.As a safe,non-invasive and non-subjectively controlled detection method,ABR detection technology has been widely used in neonatal hearing detection and intraoperative detection of patients.The most widely used method for the extraction of ABR signals in clinic is the average stacking method,However,this method is not only easy to lose effective information,but also easy to cause mental fatigue to the subjects.Therefore,it is of great significance to study the few and single extraction of ABR signals.In this paper,the waveform of the ABR signal is extracted by the combination of software and hardware.The Daubechies wavelet basis with strong partialization capacity is combined with the Coiflets wavelet basis with high frequency band division capacity to decompose noisy signals,the signal-to-noise ratio of noisy signals is improved.so the RBF neural network can extract waveform better even under the low SNR.Then the signal is input to the RBF neural network,and ABR signal waveform features can be extracted using the infinite approximation characteristics of the RBF neural network.Since the learning rate and convergence error cannot reach the optimal state simultaneously due to RBF neural network gradient fixation,the genetic algorithms are used to optimize and improve it.After that,in the paper a hardware circuit for collecting EEG signals is designed based on the STM32f407 single chip microcomputer.It mainly includes the main control module,stimulating sound drive module,signal conditioning module,A/D collection module,and communication module,etc.The software interface of the ABR signal collection are also designed.During the experimental process,the ABR signal extracted from different methods is compared.The ABR signal error obtained by the SWT-GA-RBFNN method is the smallest and the features of the signal close to the widely used clinical superposition average method,which indeed realize single extraction of ABR signals...