Introduction
Understanding emotions or objective sentiment analysis has always been a diffi cult task. Tradi- tional methods are often subjective and not always accurate, while ECG signals can provide a potential link to emotional states through changes in heart rate variability. The motivation for this project is to bridge physiological signals and emotion recognition, offering a novel yet practical approach for detecting emotions, more specifi cally, exploring how ECG data can be used for sentiment analysis. This could have valuable applications in healthcare for monitoring stress and emotional well-being and is even more useful in the society of AI, by improving HCI (human-computer interactions), systems can respond more intelligently to users’ emotional states.
This project will focus on the relationship between electrocardiac pattern with emotions and sentiments by analyzing the ECG signals recorded by Apple Watch.
Signal source
The signals are generated through the use of a specialized electrical heart sensor by operating in conjunction with the Digital Crown that contacts with my fi nger and the back crystal of the watch to record an ECG and collected from the 'Health app' of my Apple Watch Series 7 and exported from the 'Health app' on the iPhone connected to that Apple Watch. The sampling rate is 512Hz, which is shown in the exported health record. This ECG on the Apple Watch has decent accuracy, but there is still potential noise compared to medical ECG devices, especially:
- Movements: Occur due to physical movements, such as arm or wrist movement, which may cause abrupt fl uctuations or distortions in the signal.
- Contact Noise: In wearable devices, poor contact between the skin and sensors due to sweat, dry skin, or blotches can introduce noise. However, on 'Health app', the recording process will be forced to restart if dropouts happen.
- Powerline Interference: This one is rare and hence might not be considered due to technical diffi culties in this project. Under strong electromagnetic fi elds environment, like signal from nearby electrical devices, this can lead to a potential periodic noise in the signal.
Sample signal plot
Scenario
A total of 10 ECG signals will be collected under different scenarios, including during watching a horror movie, after extensive exercise, before an interview and simply resting.