Voice-based Personal Assistant

Alumnus built an innovative AI based personal assistant to help detect dangers to personal safety. The assistant engages in conversation with a person and evaluates the response to make an assessment.

A solution was needed to detect deviations from “normal” speech patterns of individuals due to the stress caused by potential threats in his / her external environment. These deviations would then be ratified to proactively identify threat situations and call for help.​

The threats can originate from natural (fire, flood) or human (assailants, fugitives) causes, or from specialized scenarios such as medical emergencies and accidents in enclosed spaces.

​Alumnus designed a voice-base Bot that interacts with users on a mobile to perform regular tasks. During the delivery of these tasks, the Bot also learns the ‘normal’ speech pattern of that user (loudness, pitch, etc.).

​In any subsequent call, if the Bot solution detects any deviation from ‘normal’ speech pattern, it switches to Incident Mode to auto-generate calls for help and, if deemed necessary, dispatch help to ensure the user is safe. Deviations can also be caused by unexpected responses (not learnt by the Bot) and from ‘hot’ words (pleas for help or cuss words, for example).