The Third Clarity Prediction Challenge (CPC3)

Speech intelligibility prediction is fundamental in the design of hearing aid algorithms to improve real-world communication outcomes for patients. The third Clarity Prediction Challenge (CPC3) builds on previous challenges by increasing the size and generalisability of the data set, allowing us to predict intelligibility in real listening environments more accurately and reliably. CPC3 was launched in March, and the deadline for submissions is 31st July, 2025. The results of this new challenge will be showcased at an International Speech Communication Association workshop, which will be an Interspeech 2025 satellite event held in Delft on 22nd August, 2025.

The challenge involves estimating the intelligibility of speech-in-noise processed by hearing aid algorithms when auditioned by listeners with hearing loss. Both intrusive (where the clean speech is available to the algorithm) and non-intrusive systems are welcomed. The ranking will be the overall performance across the evaluation data set. The training, development and evaluation data sets provide intelligibility scores from our panel of listeners with hearing loss. The training data is derived from the 1st and 2nd Clarity Enhancement Challenges (CEC1, CEC2), while development and final evaluation data come from the 3rd Clarity Enhancement Challenge (CEC3). The scenes in the training data include both simple scenes with a single interferer and more complex scenes with multiple interferers. The development and evaluation data scenes include a mixture of simulated scenes and real-world recordings. For more information, and to access our publicly available leaderboard, please visit claritychallenge.org.