Scientists from MIT in collaboration with Artificial Intelligence Lab (CSAIL) have found a new approach to recognize human emotions. Scientists used EQ-Radio device to bounce waves off subjects and calculate their breathing patterns and heart rates. EQ radio device emits and captures reflected radio frequency (RF) waves.
The data then runs through an algorithm. The algorithm was programmed to match physical factors with different emotional evidence. It then classifies the person’s emotion as one of four states: sadness, anger, pleasure, and joy.
Scientists took 12 participants. They instruct them to sit three to four feet away from the device. The researchers instructed the subjects to recall a personal experience that made them feel a certain emotion. That established a “ground truth,” or an emotion that the researchers could test to see if the machine would catch it.
The device then uses one of two methods to predict the subject’s feelings. It may either use person’s ground truth as a baseline for future predictions or the baselines gathered from the other participants’ reported emotions.
After that, each subject recalled different experiences that trigger one emotion out of the four emotions the machine could detect. The device measures heart and breathing rates within two minutes.
As the result, the device shows 87% accuracy when depends on person’s baseline and 72% accuracy with people’s emotions.