Haoshuai Zhou

Haoshuai Zhou

Not only an AI Engineer

Selected Publications

Conference Papers

Intrusive Intelligibility Prediction with ASR Encoders

Published in Interspeech, 2025

Speech foundation models have demonstrated strong performance in speech intelligibility prediction. In this paper, we present a set of practical techniques that enabled us to achieve first place in the Third Clarity Prediction Challenge.

Recommended citation: Yu, H., Zhou, H., Cao, B., Mo, C., Li, L., & Wang, S. X. (2025, August). Intrusive Intelligibility Prediction with ASR Encoders. In Clarity-2025: The 6th Clarity Workshop on Improving Speech-in-Noise for Hearing Devices (Satellite of Interspeech 2025), Delft, The Netherlands.
Download Paper

No Audiogram: Leveraging Existing Scores for Personalized Speech Intelligibility Prediction

Published in Interspeech, 2025

In this paper, instead of using audiograms, we propose a new paradigm for personalized speech intelligibility prediction, which utilizes one’s existing intelligibility scores to predict the intelligibility score on new audio clips.

Recommended citation: Zhou, H., Mo, C., Cao, B., Li, L., & Wang, S. X. (2025). No Audiogram: Leveraging Existing Scores for Personalized Speech Intelligibility Prediction. Interspeech 2025.
Download Paper

Unveiling the Best Practices for Applying Speech Foundation Models to Speech Intelligibility Prediction for Hearing-Impaired People

Published in WASPAA, 2025

Speech foundation models (SFMs) have shown strong performance across various downstream tasks, including speech intelligibility prediction. In this paper, we provide a comprehensive guide highlighting the key considerations necessary to maximize prediction performance.

Recommended citation: Zhou, H., Cao, B., Mo, C., Li, L., & Wang, S. X. (2025). Unveiling the Best Practices for Applying Speech Foundation Models to Speech Intelligibility Prediction for Hearing-Impaired People. WASPAA 2025.
Download Paper