The content of this page is only for informational purposes and is not intended, expressly or by implication, as a guarantee of employment or salary, which vary based on many factors including but not limited to education, credentials, and experience. Alliant International University explicitly makes no representations or guarantees about the accuracy of the information provided by any prospective employer or any other website. Salary information available on the internet may not reflect the typical experience of Alliant graduates. Alliant does not guarantee that any graduate will be placed with a particular employer or in any specific employment position.
Emotion Recognition is defined as the process of detecting human emotion like happy, sad, angry etc. Emotion recognition is studied in many branches of science like neuroscience, psychology, and cognitive sciences. It is also active research area in computer science. In computer science, emotion recognition leverages the research done in human-machine interaction to accurately predict the human emotion based on some kind of features. This can broadly be done in two ways. One, we can use facial features like facial expression, facial cues, hand or body gestures and voice intonation patterns for recognizing emotion. Secondly, we can use the common physiological features like electroencephalograph (EEG), electrocardiograph (ECG), pulse and breathing rate.
Compared to earlier approach, the second approach, based on physiological features, is much more robust and accurate. Even though modern computer vision techniques allow for very accurate detection of facial features, training the system on any kind of subtle emotional expression is very hard and prone to errors. On the other hand, techniques based on EEG signals are much more robust to errors, and thus give much more reliable results. SVM and KNN have been extensively used for doing emotion recognition based on EEG signals. At CSML (California School of Management and Leadership) in Alliant, we are investigating the results of an EEG based emotion recognition framework using Deep Learning. Machine Learning models are typically categorized as shallow or deep models. Shallow models include traditional ML classifiers like Support Vector Machines (SVM), k Nearest Neighbors (KNN), Single Layer Perceptrons. Deep Models are based on multi- layered neural networks like Convolutional Neural Networks (ConvNets), Recurrent Neural Networks (RNNs), Deep Belief Networks (DBNs). Out of these, ConvNets are especially popular, especially for multi-channel signals like images and medical signals. Recently, deep models have gained huge traction due to their performance gains over traditional methods (sometimes they perform 5-10x better than traditional methods). Hence, it makes sense to use Deep Learning for emotion recognition and has been done by a few researchers. But Deep Learning models also require lots of data which is unavailable in the case of emotion recognition problem. Thus, more data is needed if we intend to apply deep learning to emotion recognition, and the best way to do that is through data augmentation.
Starting Your Data Analytics Career
If the idea of emotion recognition captivates your imagination and you’d like to learn more about peripheral analytics, you’ll get a lot out of the Master of Science in Data Analytics (MSDA) degree program at Alliant International University. The MSDA program is focused on growing your skills in data science through technology and quantitative analysis; statistical modeling; critical thinking; and deep analysis, forecasting, and operational problem solving.
Available both on our San Diego campus and fully online, the MSDA data analytics course or the IST degree program teaches you valuable skills through real-world applications, and ensures that you’re fully prepared to work as a professional within organizations as the mind that can master the machine.