A Neural Network Approach for Anxiety Detection Based on ECG
Adrian Vulpe-Grigorasi, Ovidiu Grigore
Abstract
Electrocardiogram (ECG) analysis has been used with success as a method for mental stress assessment. Markers based on heart rate variability (HRV) show a correlation between HRV and emotional arousal caused by anxiety. This paper presents a deep learning approach for anxiety detection in arachnophobe individuals based on their ultra-short HRV variability measures. The results obtained indicate that 1D convolutional neural networks (CNN) trained on ECG derived features can be used for anxiety detection. Validation accuracy, precision and recall for the proposed method were respectively of 83.29%, 85% and 82%.
Topics & Concepts
Heart rate variabilityConvolutional neural networkAnxietyComputer scienceCorrelationRecallArtificial intelligenceArousalArtificial neural networkDeep learningPattern recognition (psychology)Speech recognitionPsychologyHeart rateCognitive psychologyMedicineMathematicsNeuroscienceInternal medicineBlood pressurePsychiatryGeometryHeart Rate Variability and Autonomic ControlNon-Invasive Vital Sign MonitoringEEG and Brain-Computer Interfaces