Due to the pace of modern life and the shift of nature of work from physical to cognitive, mental stress is increasing in every profession. Mental stress has now become a leading cause of work-related illness. There are numerous sedentary occupations, such as those in the IT industry, where individuals are required to work for extended periods of time, leading to stress. Working for extended periods under mental stress can increase the risk of life-threatening diseases like cardiovascular diseases, mental health disorders etc. There is, thus, a requirement for a non-obtrusive tool to detect mental stress. In this work, pulse rate variability (PRV) of 15 subjects during 5 cognitive states (relaxation, deep breathing, and during three mental tasks involving three levels of mental stress) was examined using photoplethysmography (PPG). The result of this study indicates that 18 features (9-time domain and 9 frequency domain) are statistically significant at p < 0.05 as per the Friedman test in 5 cognitive states. The machine-learning algorithm based upon a multi-layer perceptron (MLP) was able to classify with an overall accuracy of 85.1±1.1%. Classification accuracy further improved by using deep neural networks (DNN) to 91±1.1%.

由于现代生活的节奏和工作性质从体力到认知的转变，每个职业的精神压力都在增加。精神压力现已成为与工作有关的疾病的主要原因。有许多久坐不动的职业，例如在 IT 行业，人们需要长时间工作，从而导致压力。在精神压力下长时间工作会增加患心血管疾病、精神健康障碍等危及生命的疾病的风险. 因此，需要一种非侵入式工具来检测精神压力。在这项工作中，使用光电体积描记法 (PPG) 检查了 15 名受试者在 5 种认知状态（放松、深呼吸和涉及三种心理压力水平的三种心理任务期间）的脉搏率变异性 (PRV)。 这项研究的结果表明，根据 5 种认知状态下的弗里德曼检验， 18 个特征（9 个时域和 9 个频域）在p < 0.05时具有统计学意义。基于多层感知器 (MLP) 的机器学习算法能够以 85.1±1.1% 的总体准确度进行分类。通过使用深度神经网络 (DNN)，分类准确率进一步提高到 91±1.1%。