Improving emotion recognition methods using data science

Assistant professor in Applied Computer Science, Dr. Camilo Valderrama, wants to make it easier to understand brain emotional responses. He applies data science techniques, like AI and deep learning, to Electroencephalographic (EEG) signals - electrical signals captured from the scalp that reflect brain activity.

It's very hard to develop an emotion recognition method that is practical for many users because the brain of each person is different.

Dr. Camilo Valderrama

Dr. Valderrama received a Discovery Grant and a Discovery Launch Supplement from the Natural Sciences and Engineering Research Council of Canada (NSERC). A total of $157,500 from NSERC will go toward his project, Enhancing Subject-Independent Emotion Recognition Through Dynamic Brain Lateralization.

"I like to apply these computational models or mathematical models to bio-signals, or any type of data, to enhance the work of medical professionals," Dr. Valderrama said. "They can have this extra input, in addition to their knowledge and experience, to make decisions."

His NSERC-funded project will help to improve the development of computational methods for detecting emotions from EEG signals. Since the brain regulates emotion, EEG signals can help to train machine learning models to recognize emotions. However, because EEG signals vary among individuals, the accuracy of these models can drop for new users, needing recalibration. This makes the models less usable.

"It's very hard to develop an emotion recognition method that is practical for many users because the brain of each person is different," Dr. Valderrama explained. "We want something that, when a new patient comes, using the same methods, it can be used with minimum calibration."

Supporting healthcare

Having machines that can detect emotion could have applications in human-computer interaction, education, and marketing. However, Dr. Valderrama's research focuses primarily on how machine learning can be applied to clinical or medical domains in order to support healthcare.

"As people age, some areas are compressed in the brain and a person may have difficulties to express or feel emotions," he said. "People with some kind of specific emotional disorder, like bi-polar disorder, autism, or even stress, may have challenges analyzing or detecting emotions of others."

"If we know which brain areas are more relevant for processing emotions, this could be useful for psychologists or psychiatrists when trying to detect if a person is lacking this capacity for experiencing or expressing emotions."

Having a practical and effective tool, like a portable EEG cap, to identify emotional function would support medical diagnosis and treatment. It would also help medical professionals to easily track the progress of treatments.

Dr. Valderrama is conducting additional research that applies machine learning models to healthcare problems. He is currently using data science to help identify factors - like maternal and paternal age, ethnicity, and a mother's height - that may contribute to newborns having low birth weight.

"We're humans, so a lot of data collected is hard for us to analyze," he said. "But, if we put this into the machines, they can show us which factors are more relevant to the specific outcomes that you are analyzing."

Dr. Valderrama hopes data science research, like his, will give policy makers further resources to make informed decisions about public health.

Public link

Please use the above public link if you want to share this noodl on another website.