The WSJ0 Hipster Ambient Mixtures (WHAM!) dataset pairs each two-speaker mixture in the wsj0-2mix dataset with a unique noise background scene. It has an extension called WHAMR! that adds artificial reverberation to the speech signals in addition to the background noise. The noise audio was collected at various urban locations throughout the San Francisco Bay Area in late 2018. The environments primarily consist of restaurants, cafes, bars, and parks. Audio was recorded using an Apogee Sennheiser binaural microphone on a tripod between 1.0 and 1.5 meters off the ground.
WHAMR! is a dataset for noisy and reverberant speech separation. It extends WHAM! by introducing synthetic reverberation to the speech sources in addition to the existing noise. Room impulse responses were generated and convolved using pyroomacoustics. Reverberation times were chosen to approximate domestic and classroom environments (expected to be similar to the restaurants and coffee shops where the WHAM! noise was collected), and further classified as high, medium, and low reverberation based on a qualitative assessment of the mixture’s noise recording.
The DNS Challenge at INTERSPEECH 2020 intended to promote collaborative research in single-channel Speech Enhancement aimed to maximize the perceptual quality and intelligibility of the enhanced speech. The challenge evaluated the speech quality using the online subjective evaluation framework ITU-T P.808. The challenge provides large datasets for training noise suppressors. Source: Deep Noise Suppression Challenge – INTERSPEECH 2020
The EARS-Reverb dataset uses real recorded room impulse responses (RIRs) from multiple public datasets (ACE-Challenge, AIR, ARNI, BRUDEX, dEchorate, DetmoldSRIR, and Palimpsest). All RIRs are fullband, and a randomly selected channel for multi-channel recordings is used. The reverberant speech is generated by convolving the clean speech with the RIR. To avoid a time delay between the reverberant and clean speech signal caused by the direct path of the RIR, the beginning of the RIR is cut off up to the index with the highest amplitude. Only RIRs with an RT60 reverberation time that does not exceed 2 s are used. Finally, the loudness of the reverberant speech is normalized to the loudness of the clean speech using the loudness K-weighted relative to full scale (LKFS).
Noiseless reverberant dataset using the public WSJ0 corpus and simulated room impulse responses using the PyRoomAcoustics library. Used in: - Speech Enhancement and Dereverberation with Diffusion-based Generative Models, Richter et al., arXiv 2022 - StoRM: A Stochastic Regeneration Model for Speech Enhancement and Dereverberation, Lemercier et al., arXiv 2022 - Analysing Discriminative versus Diffusion-based Generative Models for Speech Restoration, Lemercier et al., ICASSP 2023