Cardiac and Respiratory analysis from photoplethysmographic (PPG) signal for stress and mental workload assessment

The World Health Organization (WHO) has included burnout in the International Classification of Diseases. Consequently, real-time monitoring of an individual’s stress levels and mental load while performing a task is crucial to assess whether the person is actually able to complete the assigned task. The analysis of physiological signals is the best strategy for real-time monitoring of mental workload, which have distinctive characteristics related to the individual’s physiological state. In particular, the cardiac and respiratory signals have characteristic features, such as heart rate (HR), heart rate variability (HRV), and respiratory rate (RR), that are highly dependent on mental workload levels.However, the necessity of acquiring each signal with a dedicated system gives rise to issues concerning the footprint, which hinders the task to be performed. In order to overcome these limitations, recent wearable devices can be utilised for the purpose of acquiring the photoplethysmographic (PPG) signal from which the extraction of both cardiac and respiratory characteristics is possible. This development has the potential to significantly reduce the user’s bulk associated with conventional acquisition systems. The aim of this study is to extract the respiratory signal from the PPG signal. This is achieved by focusing on the detection of the respiratory frequency and on the morphology of the signal itself. A novel algorithm on the Matlab® platform that employs a diversified strategy for the extraction of respiratory signal, depending on the quality of the signal, has been developed. The first step in extracting the respiratory signal involves implementating a low-pass filter with a cutoff frequency of 0,65Hz. After the filtering phase, a statistical analysis to verify the cleanliness of the signal is conducted. In the event that the filtered signal is clear, the algorithm employs a spectral analysis method in order to identify the respiratory frequency. Conversely, when noise is present, the Empirical Mode Decomposition (EMD) for signal decomposition is applied. This solution ensures enhanced computational efficiency and is particularly well-suited to a real-time context, offering a sophisticated and expeditious methodology for monitoring the respiratory signal. The efficacy of the algorithm was tested on the Capnobase data set, consisting of 42 physiological signals including respiration and PPG, acquired at rest for 8 minutes. The results are compared with the respiration rate values obtained on the same dataset using the Smart Fusion algorithm. The comparison demonstrated that the former led to superior results, especially for very noisy signals on which the latter was unable to perform. The validation of the algorithm was achieved by creating an experimental data set using the BiosignalplusX device, which is capable of acquiring the ECG, respiratory signal, and PPG signal simultaneously. Patients firstly underwent a resting phase and subsequently an N-Back test, the purpose of which was to stimulate cognitive function. The algorithm demonstrated remarkable efficacy in identifying respiratory and heart rate by directly comparing the extracted data with the acquired references. In the assessment of mental load, the extracted feature distribution were compared with the results of subjective questionnaires, which investigated the perception of cognitive effort during the test.