philgzl/mls-hq-urgent-track1

Multilingual LibriSpeech HQ (MLS-HQ) 是一个多语言语音数据集，用于URGENT 2025 Track 1挑战赛。该数据集是原始多语言LibriSpeech HQ数据的镜像版本，音频文件已从FLAC格式转换为Opus格式，以减少文件大小并加速流式传输。技术规格包括：采样率为48 kHz（从44.1 kHz重采样以支持Opus格式），单声道，Opus格式。数据集包含三个语言分区：西班牙语（150小时，36031个话语）、德语（150小时，35890个话语）和法语（150小时，36078个话语）。许可证为CC0 1.0，数据来源于HuggingFace上的kohei0209/mls_hq_urgent_track1数据集，并关联到Interspeech 2025 URGENT语音增强挑战赛的论文。Multilingual LibriSpeech HQ (MLS-HQ) is a multilingual speech dataset used in the URGENT 2025 Track 1 challenge. It is a mirror of the original Multilingual LibriSpeech HQ data, with audio files converted from FLAC to Opus format to reduce size and accelerate streaming. Technical specifications include: a sampling rate of 48 kHz (resampled from 44.1 kHz to support Opus format), mono channel, and Opus format. The dataset consists of three language splits: Spanish (150 hours, 36031 utterances), German (150 hours, 35890 utterances), and French (150 hours, 36078 utterances). It is licensed under CC0 1.0, sourced from the HuggingFace dataset kohei0209/mls_hq_urgent_track1, and associated with the paper for the Interspeech 2025 URGENT Speech Enhancement Challenge.