Public Release: 

CVSSP celebrate success in Google-sponsored audio challenge

An advanced audio classification system developed at Surrey, which can recognize individual sounds within everyday environments, has been ranked third out of 558 systems worldwide in the DCASE 2018 Challenge

University of Surrey

As part of the DCASE 2018 Challenge, the Kaggle Freesound Audio Tagging Challenge asks competitors to create a machine to identify and understand everyday sounds such as a dog barking, a telephone ringing, or a guitar strumming.

Surrey's Centre for Vision, Speech and Signal Processing (CVSSP) team went head to head with top industry and academic players in the field of acoustics. Team members Turab Iqbal, Qiuqiang Kong, Professor Mark Plumbley and Dr Wenwu Wang outperformed major international corporations and global university groups in the challenge to reach the top of the ranking list. Funding was provided by the Engineering and Physical Sciences Research Council project 'Making Sense of Sounds', which is a collaboration between the universities of Surrey and Salford.

CVSSP's general purpose audio tagging system uses artificial intelligence (AI) to simulate the auditory function of a human being. This improves a machine's situational awareness of sounds, enabling better decision making. Such technology will be incredibly useful for many applications including robotics, assisted living, security surveillance, home smart devices, smart cities sensors, environmental monitoring, and situational awareness in defence.

Dr Wang, Reader in Signal Processing, said: "This is an outstanding achievement following our success on the DCASE 2017 Challenge, for which we also topped the ranking list. This further confirms our world-leading research in sound recognition, in particular sound event detection, scene classification and audio tagging."

Professor Adrian Hilton, Head of CVSSP, added: "This is an excellent achievement for the team, demonstrating that the centre has world-leading AI and Machine Perception research in both audio and visual recognition. Our challenge over the next decade is to bring these technologies together to enable future intelligent systems for healthcare, robotics, automotive and entertainment sectors."

CVSSP's results will be presented at the DCASE 2018 Workshop to be held in Woking on 19-20 November.

This is the fourth DCASE challenge since the competition launched in 2013; it was organised by Google and Universitat Pompeu Fabra. Competitors included Tampere University of Technology, New York University and Korean Advanced Institute of Science and Technology.

Read more about CVSSP's success in the DCASE 2017 Challenge last year and the Google Landmark Retrieval Challenge earlier this year.

Explore CVSSP's Electrical and Electronic Engineering programmes.


Disclaimer: AAAS and EurekAlert! are not responsible for the accuracy of news releases posted to EurekAlert! by contributing institutions or for the use of any information through the EurekAlert system.