freococo/Google_Myanmar_ASR

**Google Myanmar ASR数据集**是基于**缅甸语语音语料库**的，最初由Google发布。该数据集包含音频文件及其对应的转录文本，主要用于构建自动语音识别（ASR）模型。数据集的主要特点包括：语言为缅甸语，所有音频文件重采样至16 kHz，提取了80 mel bins和3000帧的梅尔频谱特征。数据集结构包括训练和测试数据，分别存储在`/train`和`/test`文件夹中，并提供了预计算的特征和元数据文件。预处理步骤包括音频重采样、特征提取和元数据准备。

The **Google Myanmar ASR Dataset** is based on the **Burmese Speech Corpus**, originally published by Google. It consists of audio files and their corresponding transcriptions. The dataset is primarily aimed at building Automatic Speech Recognition (ASR) models. Key highlights include: Language is Myanmar (Burmese), all audio files are resampled to 16 kHz, and Mel-spectrogram features are extracted with 80 mel bins and 3000 frames. The dataset structure includes training and testing data, stored in `/train` and `/test` folders respectively, and provides precomputed features and metadata files. Preprocessing steps include audio resampling, feature extraction, and metadata preparation.