Predicting Premature Video Skipping and Viewer Interest from EEG Recordings

Abstract

Brain–computer interfacing has enjoyed growing attention, not only due to the stunning demonstrations with severely disabled patients, but also the advent of economically viable solutions in areas such as neuromarketing, mental state monitoring, and future human–machine interaction. An interesting case, at least for neuromarketers, is to monitor the customer’s mental state in response to watching a commercial. In this paper, as a novelty, we propose a method to predict from electroencephalography (EEG) recordings whether individuals decide to skip watching a video trailer. Based on multiscale sample entropy and signal power, indices were computed that gauge the viewer’s engagement and emotional affect. We then trained a support vector machine (SVM), a k-nearest neighbor (kNN), and a random forest (RF) classifier to predict whether the viewer declares interest in watching the video and whether he/she decides to skip it prematurely. Our model achieved an average single-subject classification accuracy of 75.803% for skipping and 73.3% for viewer interest for the SVM, 82.223% for skipping and 78.333% for viewer interest for the kNN, and 80.003% for skipping and 75.555% for interest for the RF. We conclude that EEG can provide indications of viewer interest and skipping behavior and provide directions for future research.

A Systematic Review for Cognitive State-Based QoE/UX Evaluation

“…Libert and Van Hulle [ 40 ] evaluated interest in videos using kNN with EEG characteristics based on entropy and indices of engagement, valence, and activation calculated considering the power of different frequency bands.…”

“…In [ 40 ], the experiment was carried out with few participants, and it consisted of evaluating interest or omission in observing a set of 45 videos presented one-by-one in three blocks to avoid fatigue.…”

“…Experiments with different signals and number of participants were identified: EEG data from 4 participants [ 40 ], ET data from up to 136 participants [ 39 ], or acquiring data from various signals from up to 61 participants [ 34 ]. Figure 3 shows the distribution of the number of participants in experiments that collected EEG, ECG, or GSR data and of ET with medians of 20, 42.5, 24, and 33.5, respectively.…”

“…Libert and Van Hulle [40] evaluated interest in videos using kNN with EEG characteristics based on entropy and indices of engagement, valence, and activation calculated considering the power of different frequency bands. One of the investigations [34] presented an experiment with free browsing of a fictitious website, detecting active or transition windows based on gaze fixations at predefined areas of interest.…”

Neuromarketing as an Emotional Connection Tool Between Organizations and Audiences in Social Networks. A Theoretical Review

“…This study is a pioneer in the use of neurophysiological methods to predict advertising success in a digital context. Likewise, some researchers have shown that the electroencephalography (EGG) technique can provide indications about a subject’s interest in watching a video or the possibility of closing and skipping it without seeing it ( Libert and Van Hulle, 2019 ).…”