“I want to create alarms to help reduce traffic accidents by designing emotion recognition models,”, said Mr. Tran Bao Toan, from the DTU Center of Software Engineering (CSE), who won first prize in the Emotion Recognition Challenge contest, with his “Emotion Recognition Machine Learning from Voice” application, at the ERC 2019 competition.
The ERC2019 topic, entitled “Pioneering solutions for Emotion recognition”, concentrated on a variety of text, image, sound and video data types and was the first contest based on sound-based emotion recognition to be held in Vietnam. The contest also aims to identify and promote talented researchers from Vietnam and elsewhere.
Mr. Tran Bao Toan graduated from DTU in 2017 and then started working at DTU CSE. He said: “I love technology and I dreamed of becoming a programmer when I was young. The environment at DTU CSE is ideally suited to my interests and qualifications. By working here, I want to broaden my abilities by participating these types of intensive competitions”.
Mr. Tran Bao Toan (middle). Photo. TH
Initially, entrants received a training dataset with 7,442 samples of wave-label file pairs:
• Wave: spoken data in PCM 16 kHz, 16 bit, mono format
• Label: a text file with six emotion tag wave files depicting happiness, sadness, anger, fear, disgust and neutrality
The leading nine models were selected to compete in the finals on December 15, 2019 when the finalists each presented their work and the judges tested the systems privately.
Mr. Tran Bao Toan then designed his impressive model. “I created my machine learning model using an artificial neural network, which was able to distinguish between the six human emotions through voice recordings”, he explained. “Then I uploaded my model to compete with the other contestants. I was elated that I won first prize.”
Mr. Tran Bao Toan expects that his model can be applied in many different fields. For example, it can recognize the emotional state of drivers, and fatigue, drowsiness, drunkenness can all be detected and an alarm sounded. In healthcare, the application can be used to detect medical symptoms by vocal analysis and help doctors draw up improved treatment strategies. He also wants to upgrade his model so that it can be more widely used in our daily lives.
(Media Center)