Study of Evoked Emotional Responses

Millions of people are affected by mental health conditions every year. Approximately one in five adults in the U.S. experiences at least one mental illness episode every year. Approximately 16 million people have at least one major depressive episode in one year. Researchers have identified that Millennial age-group is at highest risk of developing anxiety, depression, and thoughts of suicide than any other generation. These conditions are attributed by many reasons like increasing competition, less communication, less real-world attachment, work pressure, etc.

Considering the increasing level of issues related to mental health, compelled to think about understanding emotions and associated health impact. Popular methods of emotion detection are by detecting facial expression but to detect real-time emotion, biophysical signals are the best tool as per literature. Also, by considering the future possibility of product form of EMO-001, emotion detection through bio-signals proved to be the correct approach. The intent of the research is to classify the evoked emotions based on bio-signals in support of the product EMO-001 device using photo plethysmography (PPG), Galvanic Skin Response (GSR) and Electrocardiogram (ECG). Identification of emotions from bio-signals, and development of classification algorithm helps the product to identify the emotional state of the person real-time and eventually helps to track the emotional state and suggest an activity which can lift the emotional well being of a person.

Major six emotions, amusement, sad, disgust, scared, surprised, thrill were considered to collect bio signal. Bio signals for the neutral state were also captured to create a baseline for identifying changes in the bio signals with changing emotions.

To find the gender balance in subjects, a similar ratio of male and female considered. Targeting Millennial age-group for this product, we consider people from the age of 18 to 33 years with a minimum bachelor degree qualification and those who are employed. We considered people who understand English and Hindi are considered as subject as video content were in these languages only.

Exploring different methods of emotion elicitation, an audio visual method was considered for this trial, as it is the most effective methods described in different previous literature and through practical scenario. More than 1,000 Video clips from open platform were seen and analysed for five different emotions. Out of which 100 video (20 per emotion) clips were scrutinized by three level checks, selected video clips were also validated by internal blind check to figure out what exact emotion has been elicited from each video. From these identified clips, different set of clips had been prepared. We could identify that video clips are less effective to elicit the anger; hence interaction method was used for it.

Internal validation of data capturing sensors and assembly done to ensure the bio signals obtained are of good quality. For the collection of bio signals, subjects were asked to sit in a proper position, and sensors were placed at particular place of the body.

When the subject was ready for the participation, initially for 30 seconds neutral pictures (e.g. nature sites) were shown to neutralize his/her emotions. After that starting with any of the emotion, three randomly selected videos of a particular emotion from data bank was shown, each followed by a neutral clipping to neutralize the effect of the earlier video. In parallel of videos, the bio-signal data was collected from the sensors. After each video, the subject was asked to fill up a feedback form to rate his emotions. After finishing the signal collection of five different emotions and neutral in between, interaction with subject initiated where the subject was asked about their general background and slowly turned to some sensitive topic which helped to elicit the anger emotion.

The trial had completely non-invasive intervention, where participant need to sit for two hours to watch the videos. Collected data would further be analysed and used for development of emotion classification algorithm.