That which you feel
We measure emotions. We can’t stress it enough: emotions are at the core of a person, driving all their actions, which is why we must take them into account. But how do we measure emotions anonymously? That’s where our technology comes into play.
Emotion AI, also known as affective computing, is a rapidly growing branch of Artificial Intelligence that enables computers to analyze and understand non-verbal human cues to infer the emotions being expressed. Among all high-level vision tasks, visual emotion analysis is one of the most challenging due to the gap between low-level pixels and high-level emotions.
However, achieving this requires not only technology but also psychology. Emotional expressions, both facial and bodily, are universal. An emotional expression doesn’t uniquely identify a person but rather an emotion. Thus, emotions can be identified using taxonomy with convolutional neural networks (CNNs)
The smiles data warehouse Golineuro AI incorporates algorithms for recognizing hundreds of emotions, ranging from boredom to embarrassment, including indifference, curiosity, or empathetic pain. As mentioned in this blog entry, to identify emotions, we must not only consider facial expressions but also overall verbal language.
In this way, an algorithm specialized in emotion detection detects the shape of a smile and squinted eyes but also the body tension typical of this emotion. Taxonomically, it would add a smile (only the data, not the image), a squinted eye expression, and muscle tension to the data warehouse.
By increasing these three counters, it would increase the joy emotion counter by one, as this emotion would be composed of the sum of these detections.
Thus, one can access the number of smiles, muscle tension, squinted eyes, or joys. However, it’s impossible to know which individuals expressed them as there is no identifying information. The key to emotion detection lies in the set of shape recognition algorithms related to emotions, the created taxonomy, the set of algorithms specialized in different camera angles and lighting conditions, and the platform’s knowledge base, “fed” by over eight years of work in various real-world environments with millions of people.
