August 30th, 2023: The MISP 2023 challenge has been accepted by ICASSP 2024 Signal Processing Grand Challenge (SPGC). Please refer to https://2024.ieeeicassp.org/sp-grand-challenges for more details of ICASSP 2024 SPGC.
September 1st, 2023: The download link of training set and development set has been sent to the registered participants via email.
Following the success of the first and second MISP challenge, we are pleased to announce the third MISP challenge. In light of the contributions of the front-end system and the finding that visual cues can aid human speech perception, the focus of the MISP 2023 challenge is on the audio-visual target speaker extraction (AVTSE) problem, which aims to extract the target speaker’s speech from mixtures containing various speakers and background noise. With both audio and visual data provided, the challenge considers the problem of audio-visual distant multi-microphone signal processing in everyday home TV scenarios, where several people are chatting while watching TV in the living room. The strong background noise, high overlaps ratios, and possible blurry videos are the challenges in MISP dataset. We warmly invite researchers from both academia and industry to participate in our challenge for promoting speech processing research using multi-modal information to cross the practical threshold of realistic applications in challenging scenarios.
MISP 2023 challenge features one task:
Audio-Visual Target Speaker Extraction
On this web site you will find everything you need to get started, including,
You can find a clear description of the task setting, dataset, and baseline in the overview paper
For additional information, please email us at mispchallenge@gmail.com.