Impact of experience sampling methods on tap pattern based emotion recognition
Smartphone based emotion recognition uses predictive modeling to recognize user's mental states. In predictive modeling, determining ground truth plays a crucial role in labeling and training the model. Experience Sampling Method (ESM) is widely used in behavioral science to gather user responses about mental states. Smartphones equipped with sensors provide new avenues to design Experience Sampling Methods. Sensors provide multiple contexts that can be used to trigger collection of user responses. However, subsampling of sensor data can bias the inference drawn from trigger based ESM. We investigate whether continuous sensor data simplify the design of ESM. We use the typing pattern of users on smartphone as the context that can trigger response collection. We compare the context based and time based ESM designs to determine the impact of ESM strategies on emotion modeling. The results indicate how different ESM designs compare against each other.
|Emotion detection, User interaction, Data collection, Tap sensing, Experience sampling|
|ACM International Joint Conference on Pervasive and Ubiquitous Computing and ACM International Symposium on Wearable Computers|
Ghosh, S, Chauhan, V, Ganguly, N, Mitra, B, & De, P. (2015). Impact of experience sampling methods on tap pattern based emotion recognition. In Adjunct Proceedings of the ACM International Joint Conference on Pervasive and Ubiquitous Computing and Proceedings of the ACM International Symposium on Wearable Computers (pp. 713–722). doi:10.1145/2800835.2804396