How Silent Talker Classifies Behaviour?
Our technology processes a video stream received from a camera or a recording made earlier to make a classification (e.g. if an interviewee is truthful or deceptive). It uses large numbers of AI components to find useful features in the image and their states. The states of the features become the input channels for classification. A change in a single channel is a micro-gesture and combinations of these over some time period are analysed by a final AI component to classify the mental state of the person being observed.
This was inspired by the pioneering work of David Efron in the 1940s, who developed a manual method for psychologists to analyse film recordings. These were viewed repeatedly, looking for a specific behaviour such as a blink, one frame at a time. Silent Talker extracts similar information, but in larger amounts and faster than a human brain can handle. It also replaces the human analyst’s subjective judgment about these features with a final AI classifier which is objective. This is because it has learned from examples only and avoids problems like prejudice and confirmation bias.
Silent Talker has distinct advantages over other detection of deception techniques, and it can also provide a statistical measure of its confidence in each classification.