Randy E.S. Harnarinesingh and Chanan S. Syan
Emotion Recognition, EEG, Market Research
Opinion polls are used extensively in market research to gather information about general consumer product preferences. Some of the existing methods of opinion polling include contact interviews and paper-based questionnaires. These polling methods are however time consuming and possible erroneous since there exists a possibility of false answers provided by the interviewees. It would be advantageous to identify subject opinion automatically without the use of an overt answering mechanism such as form completion or personal interviews. This paper evaluates the feasibility of a novel Electroencephalography(EEG)-based system for the identification of user product preferences on female shoes. The system identifies subject preference automatically without the need for questionnaires. Five (5) female volunteers between the ages of 18-26 participated in a stimulus presentation paradigm developed in MATLAB 7.8.0.347 (Natick, MA). They were presented with images of female shoes on a computer screen and their electroencephalographic derivations were concurrently recorded. Differential frequency features were extracted from the recorded EEG and were used to represent changes in the subject’s emotional state during product presentation. The system relied on the emotion inducing effect of picture presentation to identify product preference. A Linear Discriminant Analysis (LDA) classifier was used to identify changes in a(8-12Hz) and ß(18-24Hz) frequency components due to product presentation in order to identify subject opinion. The average system accuracy for the five (5) subjects was 61.33% for the EEG-based inference of like/dislike opinions. The obtained system accuracy is greater than chance (50%), however the accuracy is not high enough for the system to be practically implemented. However, the system is in the preliminary stages of development and as such is likely to undergo further developments. Consequently, the system could be augmented in the future by using additional physiological variables to aid in the recognition of emotional changes that accompany product presentation. Additionally, the stimulus presentation paradigm could be redesigned to avert a subject boredom/stimulus redundancy issue which arose during EEG data collection.
Important Links:
Go Back
