A method of unsupervised machine learning based on self-organizing map for BCI

Brain computer interface (BCI) is a technology that controls computers or machines using the thoughts or intentions of a person. EEG signal measured from the human scalp is reflected with thoughts and intentions of a person, and when using the signal processing technologies such as machine learning or pattern recognition, intentions of such users can be interpreted. This study has proposed an autonomous machine learning method applicable to BCI based on Kohonen’s self-organizing map which is one of the representative methods of unsupervised learning. To achieve this, learning area adjustment method and autonomous machine learning rules using interactive functions were proposed. Learning area adjustment and machine learning have used the side control effects according to the interactive functions based on Kohonen’s self-organizing map. After determining the winning neuron, the connection strength of neuron was adjusted according to the learning rules. As the learning area gradually decreased according to the increase in the number of learning, the flow towards the input of weighted value of output layer neuron was mitigated and an autonomous machine learning that can complete the learning as the network reaches the equilibrium state was proposed. This BCI technology that connects the brain and machines is predicted to be applicable to a variety of BCI applications in that unlike the existing manual mechanism, it is applied with person’s brainwaves.