Systematic Modeling Of Brainstrip Electrical Waves Between Computer And Brain

The brain signal is seen as a wave consisting of different frequencies, which we are not able to separate the frequencies that make up it with our eyes. Each frequency spectrum of brain waves has different origins in the brain, and therefore, by recording from each point of the skull, we see a different wave, which will be different with the change of the person's state and condition, as well as the location of the electrode. The entire range of brain waves recorded by EEG is considered from about ۰ to ۱۰۰ Hz. that the researchers have categorized brain waves as below for better categorization and attribution of different functions. (Please note that the frequency numbers listed may differ in different places due to different categories and research. Brain signals (EEG) have important vital applications in various fields of medicine, etc. Despite the high efficiency of EEG signals, the presence of disturbing artifact signals is an inevitable problem. Artifacts are unwanted disturbances that are mainly caused by human activities such as muscle activity, blinking, and environmental interference such as city electricity fluctuations and can cause changes in shape and ambiguity in EEG signals. We are trying to remove some interferences by applying deep networks of stack sparse autoencoder and wavelet transform. Our proposed mechanism is as follows: in the first step, it extracts healthy signals from the existing dataset, and then by adding random noises with different SNRs in the frequency range of muscle and eye artifacts, noise is added to the healthy signals. We create and train a deep network based on Stacked Sparse Autoencoders based on greedy algorithms layer by layer. In this step, the wavelet conversion of the noisy signal is considered as the input of this network and the clean signal as its output. The advantages of the frequency domain and the wavelet transformation are used to increase the power of the network in detecting and removing noises, each of which has a specific frequency range.

کلیدواژه ها:

Electric waves ، artifact ، sparse autoencoder ، violet transform ، electroencephalogram