Abstract

Acoustic event detection (AED) refers to the task of detecting target events in audios. It provides artificial intelligence systems the capability to better understand surrounding environments. In this talk, I will discuss our team’s work on Alexa Guard, the first AED powered feature for Amazon Alexa. I will also present our research efforts on improving the accuracy and efficiency for AED models, via better model architecture, data information leveraging, and model compression etc.

Biography

Ming Sun is leading acoustic event detection (AED) modeling team at Amazon Alexa Speech. His team build production AED models for Alexa, including Alexa Guard, which detects smoke and carbon monoxide alarms and glass breaking sounds. Previously, Ming Sun worked on Alexa wake word detection, building production wake word systems for launching Alexa/Echo and 3rd Party devices. Prior to joining Alexa, Ming Sun obtained his Ph.D. at the Department of Electrical and Computer Engineering (ECE) and Center for Language and Speech Processing (CLSP) at the Johns Hopkins University.