monster-monash/AudioMNIST-DS

AudioMNIST-DS数据集是MONSTER项目的一部分，包含30,000个音频样本，每个样本时长约为1秒，采样频率为4 kHz，样本长度为4,000。这个数据集是由60位不同年龄和性别的说话者录制的数字0到9的音频，每位说话者为每个数字录制了50个样本。数据集的目的是对音频中的数字进行分类，共有10个类别。该数据集已根据说话者进行交叉验证分折，确保同一说话者的录音不会同时出现在训练集和验证集中。数据集遵循MIT许可。

AudioMNIST-DS dataset is part of the MONSTER project, containing 30,000 audio samples, each with a duration of about 1 second, a sampling frequency of 4 kHz, and a sample length of 4,000. This dataset consists of audio recordings of digits 0 to 9 by 60 speakers of different ages and genders, with 50 samples per digit per speaker. The dataset is intended for audio classification tasks, with a total of 10 classes representing the digits. The dataset has been split into cross-validation folds based on speakers to ensure that recordings from the same speaker do not appear in both the training and validation sets. The dataset is licensed under the MIT License.