2 posts tagged with "CPC3"

• 3rd Clarity Prediction Challenge (CPC-3)
• Launch: March 17th; Submission: July 31st, 2025
• Workshop: 22nd August 2025, an INTERSPEECH satellite workshop
• https://claritychallenge.org/

Dear colleague,

It gives us great pleasure to announce the launch of the 3rd Clarity Speech Separation and Recognition Challenge (CPC3).

The Challenge

To improve hearing enhancement technologies, such as hearing aids and hearable devices, we need reliable methods for automatically assessing the speech intelligibility of audio signals.

In recent years, we have organised the CPC1 and CPC2 Challenges to promote the development of such prediction systems. We are now launching the third round of this challenge, building on previous efforts by incorporating a more extensive and diverse set of listener data for training and evaluation.

What will be provided

• Hearing aid outputs: Audio produced by various simulated hearing aids while processing speech in noisy environments.
• Clean speech reference signals
• Listening Test Results: This includes transcripts and intelligibility scores from tests in which hearing-impaired listeners were asked to repeat what they heard.
• Listener Characteristics: Information regarding the severity of hearing impairment for each listener.
• Software Tools: Tools include a baseline system based on HASPI scores.

It gives us great pleasure to pre-announce the 3rd Clarity Speech Separation and Recognition Challenge (CPC-3) that will launch in March 2025.

The Challenge

To develop better hearing enhancement technologies, including hearing aids and hearable devices, we need reliable methods to automatically evaluate the speech intelligibility of audio signals. This requires a predictive model that takes as input both the audio produced by a hearing aid and listener characteristics (e.g., their audiogram) and estimates the speech intelligibility score that the listener would achieve in a listening test.

In recent years, we have run the CPC1 Challenge and CPC2 Challenge to advance such models. We are now launching the third round of this challenge, which builds on previous efforts by incorporating a larger and more diverse set of listener data for training and evaluation.

What will be provided

• Audio produced by a variety of (simulated) hearing aids for speech-in-noise;
• The corresponding clean reference signals (the original speech);
• Characteristics of the listeners (pure tone audiograms, etc);
• The measured speech intelligibility scores from listening tests, where hearing-impaired listeners were asked to say what they heard after listening to the hearing aid processed signals.
• Software tools including a baseline system based on HASPI scores.

For further details see https://claritychallenge.org/docs/cpc3/cpc3_intro