Predicting human response to memory retrieval, vigilance and task novelty using GSR and artificial neural networks

Priyadarshini R Pennathur, University of Texas at El Paso


Human operators continue to be subject to workload, and predicting their load levels becomes important in time and safety critical operations. This study is a pilot research investigating the effects of short-term memory retrieval, cognitive set shifting, and visual monitoring on the electrodermal response of humans, since electrodermal activity is one of the first responses in critical scenarios. 9 participants, 8 males and 1 female undergraduate student (Mean age = 22 years) from UTEP Industrial Engineering department participated in the study. Sternberg short-term working memory task, Wisconsin card sorting task and a standard tracking task respectively were used as the experimental tasks for each of the independent variables. Biosemi and Brain Products system was used for data acquisition and analysis. A linear mixed model fit to the data indicated that short-term memory retrieval (with amplitude as dependent variable Mean = 0.012μS, SD = 0.141μS, p = 0.022), (with reaction time as the dependent variable Mean = 1136.8030μS, SD = 472.705μS, p = 0.001) was significant. Significant correlations (p = 0.022) were obtained between GSR amplitude and reaction time. Visual monitoring significantly affected the electrodermal response (Mean = 0.0975μS, SD = 0.09290μS; p = 0.000). Electrodermal activity was also significantly affected by task novelty (Mean = -2.65079E-4μS, SD = 0.025μS, p = 0.001). Electrodermal responses were integrated into an artificial neural network for classification of load levels. Neuroshell 2 software was used for modeling the neural architecture. Artificial neural network classified the electrodermal responses into task load levels namely, low, medium and high.

Subject Area

Industrial engineering|Geotechnology

Recommended Citation

Pennathur, Priyadarshini R, "Predicting human response to memory retrieval, vigilance and task novelty using GSR and artificial neural networks" (2005). ETD Collection for University of Texas, El Paso. AAI1430973.