Recognition of preferences with analysis of EEG signals for neuromarketing applications

Neuromarketing is a blend of neuroscience and marketing. It is important because it provides deeper insights into consumer behavior by analyzing how the brain responds to marketing stimuli. This approach goes beyond traditional methods such as surveys by uncovering subconscious preferences and emotions. Neuromarketing can help create more effective marketing strategies, improve product design, improve customer experiences and deliver a competitive advantage. It is very valuable for gaining a more accurate understanding of what really influences consumer decisions. In this thesis study, it is aimed to develop a liking estimation system based on electroencephalogram (EEG) signals to be used in neuromarketing applications. The subjects were shown advertising videos for 2 different products. The EEG data recorded while watching the video will be processed to determine whether the consumer liked a product or not. Neuromarketing has become a very current and important research topic in the marketing and advertising industry. In our study, the liking status of the subjects who were shown videos and images was determined by using 32-channel EEG signals and electrodes placed on the scalp with the help of an EEG cap. After watching the advertising videos, a survey was administered to the subjects and they were asked to vote on the degree of liking of different images of the products (for example, the front view of the car, the front console, the side view, the rear view, the taillight, the logo and the front grille). 1) From the recorded EEG signals, signals corresponding to different video regions were segmented. a) It is decomposed into intrinsic mode functions (IMF) by signal analysis methods such as EMD (Empirical Mode Decomposition), EEMD (Ensemble Empirical Mode Decomposition), MEMD (Multivariate Empirical Mode Decomposition). b) Various attributes were obtained from these IMFs. [1-5]. c) The obtained features were classified with Machine Learning methods. 2) These recorded EEG signals were analyzed with time-frequency (TF) analysis methods (wavelet transform, short-term Fourier transform) and TF energy distributions were obtained as color images. These images were trained and classified by deep learning networks such as convolutional neural network (CNN) and liking estimation results were developed.