Cognitive assistant for real-time emotion detection from human speech

A method comprising:

• i) obtaining audio signal from an audio receiver;
• ii) processing the audio signal to detect if human speech is present;
• iii) responsive to the audio signal containing human speech, running a machine learning based classifier to analyze the speech audio signal and output an emotion detected in speech, wherein the detected emotion is one of calm, happy, sad, angry, fearful, surprise, and disgust;
• v) sending the emotion detected to a notification system;
• vi) looping back to i),
• wherein the machine learning classifier is periodically trained externally based on labelled audio sample data and updated in the system, and wherein the method further comprises
• receiving feedback from a user that the detected emotion was incorrect or unknown, and
• processing the feedback for the labelled audio sample data.

A non-transitory computer-readable medium comprising instructions that, when executed, cause a processing system to perform steps of:

• i) obtaining audio signal from an audio receiver;
• ii) processing the audio signal to detect if human speech is present;
• iii) responsive to the audio signal containing human speech, running a machine learning based classifier to analyze the speech audio signal and output an emotion detected in speech, wherein the detected emotion is one of calm, happy, sad, angry, fearful, surprise, and disgust;
• v) sending the emotion detected to a notification system;
• vi) looping back to i),
• wherein the machine learning classifier is periodically trained externally based on labelled audio sample data and updated in the system, and wherein the steps further comprise
• receiving feedback from a user that the detected emotion was incorrect or unknown, and
• processing the feedback for the labelled audio sample data.