Blog Digital Media Evidence

Anger and disgust: emotion recognition demo

The police investigation into Dana’s death relied heavily on the assumptions and subjective judgments of the detectives in the case. They looked at surveillance videos and saw nothing that raised red flags for them. They assumed that all was well and, therefore, that no murder had occurred. When I voiced my subjective judgments — pointing out that there was a conflict between Dana and Huck on the night before Dana was found fatally injured — investigators dismissed my observations as biased.

Emotion recognition software probably wasn’t as widely available then as it is today. One app called Face++ scans photos for faces, reads cues in facial expressions, and — based on data — analyzes the emotions expressed in the faces. Granted, algorithms are not immune to bias. But when applied to images of Dana and Huck, perhaps the software is less biased than the LBPD or me.

According to Face++, Dana was angry that night. Huck’s facial expression revealed strong indicators of disgust when he looked at Dana. As I told investigators, all was not well, and Face++ seems to back me up. Here’s a short video in which I use Face++ to analyze images from that night:

One reply on “Anger and disgust: emotion recognition demo”

[…] case involved extreme domestic violence — a woman murdered in her own home by her husband. Signs of tension and conflict can be seen in surveillance video taken inside their home. Johnson saw it and wrongly concluded that the video showed no issues. I told the police that my […]

Comments are closed.