AI-edge based voice responsive smart headphone for user context-awarenes

Abstract

Recent developments in the areas of portable devices have taken enormous developments, especially the audio/music information delivery systems like headphones either using wired or wireless connectivity with mobile/portable devices. Usage of such portable devices with connected headphones is increasing day by day with user mobility in almost all the human age groups and sectors like traveling, workplaces, home places, walk paths, etc. This usage has an issue with the user interaction with the external world and also the contextual situations, actions and responses. Such issues lead to non-involvement and inconvenience or total destruction to the user. This depends on the surrounding situation and contextual hazardous events that may lead to accidents or life-threatening injuries. In order to handle such issues, we have designed an edge-based AI Voice responsive smart headphone for context awareness and alert system to the user. This model uses an edge-based machine learning Long Short Term Memory (LSTM) algorithm for voice recognition, runs either in the headphone or the device to which it is connected. Here the voice/sound recognition and alerting the user, can handle many contextual situations like normal interactions by name, hazardous scary situational sounds either by a bystander, or external sounds. Prototype models are developed for some of the voice records/ sounds using tensor flow ML-based algorithms over low footprint devices like mobile phones or STM 32 microcontrollers. Tested and validated for some of the voice/sound scenarios.