release testset

This release contains the testset release and challenge result for the CCF AATC 2025 Speech Restoration Challenge.

Technical Report

For detailed information on dataset construction, benchmarking methodology, and a systematic analysis of the participating systems, please refer to our retrospective paper: The CCF AATC 2025 Speech Restoration Challenge: A Retrospective

Testset Description

The testset consists of paired audio files, totaling 300 pairs. The corresponding degradation types are divided into three categories: (1) Acoustic Degradation (150 files); (2) Codec Distortion (50 files) across various bitrates; and (3) Secondary Artifacts (100 files)

File structure:

clean/
├── 11.wav
├── 22.wav
├── ...
└── 99.wav
noisy/
├── 11.wav
├── 22.wav
├── ...
└── 99.wav
metadata.csv

the metadata.csv file contains the distortion type of the testset.

file_name,type
11.wav,Acoustic Degradation
22.wav,Codec Distortion
33.wav,Secondary Processing Artifacts