New mood-detecting sensors could improve AI

sensors

Left to right: PhD student Fadel Adib, PhD student Mingmin Zhao, and Professor Dina Katabi pose with their EQ-Radio device, which can detect emotion using wireless signals. Image: Jason Dorfman/ MIT CSAIL

Researchers from MIT have developed a device that utilises wireless signals to detect human emotions. Dubbed EQ-Radio, the technology works by measuring subtle changes in breathing and heart rhythms (without the use of on-body sensors) and is 87 per cent accurate at detecting if a person is excited, happy, angry or sad, according to the researchers. The project’s leader, Professor Dina Katabi, sees the technology being used for entertainment, consumer behaviour and health care.

“EQ-Radio’s system architecture has three components,” reads the research paper, co-authored by Katabi and PhD students Mingmin Zhao and Fadel Adib.

“The first component is an FMCW radio that transmits RF signals and receives their reflections. The radio leverages the approach to zoom in on human reflections and ignore reflections from other objects in the scene. Next, the resulting RF signal is passed to the beat extraction algorithm. The algorithm returns a series of signal segments that correspond to the individual heartbeats.

“Finally, the heartbeats – along with the captured breathing patterns from RF reflections – are passed to an emotion classification sub-system as if they were extracted from an ECG monitor. The emotion classification sub-system computes heartbeat-based and respiration-based features recommended in the literature and uses an SVM classifier to differentiate among various emotional states.”

sensors
Image: MIT CSAIL

 

Unlike other emotion-detecting technologies, EQ-Radio does not utilise audiovisual cues or on-body sensors. Facial expressions can be highly unreliable, according to the researchers, and ECG monitors are inconvenient to wear and become less accurate as they change position over time.

In the experiments for EQ-Radio, human subjects used stimuli including music and videos to recall memories that evoked one of the four emotions, as well as a baseline of no emotion. EQ-Radio was able to determine the correct emotion 87 per cent of the time.

The technology eclipses Microsoft’s vision-based Emotion API, which focuses on facial expressions and when tested by the researchers, had an accuracy rate of only 40 per cent. The researchers found that EQ-Radio was more accurate in detecting joy, sadness and anger. The two systems performed similarly in detecting neutral emotions, however.

According to Katabi, this work shows that “wireless signals can capture information about human behaviour that is not always visible to the naked eye”. The device could be embedded in future technologies to help monitor and diagnose conditions such as depression and anxiety, she added.

“We view this work as the next step in trying to develop computers that can understand us better at an emotional level and potentially interact with us similarly to how we interact with other human beings,” said Zhao.