TAPS (Throat and Acoustic Paired Speech Dataset)
收藏资源简介:
TAPS数据集是由韩国浦项科技大学创建的一组喉部麦克风和声学麦克风配对语音的集合,旨在为深度学习基础的语音增强研究提供标准化数据集。该数据集包含60位韩国本地说话者使用喉部和声学麦克风同时录制的6000对语句。数据集分为训练集、验证集和测试集,分别包含4000、1000和1000对语句。TAPS数据集可用于提高语音质量和恢复语音内容,有助于喉部麦克风在极端噪声环境中的实际应用。
The TAPS dataset is a standardized collection of paired speech data from throat microphones and acoustic microphones, developed by Pohang University of Science and Technology (South Korea), with the aim of providing a standardized dataset for deep learning-based speech enhancement research. This dataset includes 6000 pairs of utterances simultaneously recorded by 60 local Korean speakers using both throat and acoustic microphones. The dataset is divided into training, validation, and test sets, which contain 4000, 1000, and 1000 pairs of utterances respectively. The TAPS dataset can be utilized to enhance speech quality and recover speech content, facilitating the practical application of throat microphones in extremely noisy environments.

- 1TAPS: Throat and Acoustic Paired Speech Dataset for Deep Learning-Based Speech Enhancement韩国浦项科技大学 · 2025年



