TIMIT-TTS

TIMIT-TTS是一个由米兰理工大学和德雷塞尔大学合作创建的合成语音数据集，包含近80000条使用最先进的TTS技术生成的语音数据。该数据集旨在支持多模态合成媒体检测的研究，可以单独使用或与DeepfakeTIMIT和VidTIMIT视频数据集结合使用，进行多模态研究。TIMIT-TTS数据集的创建过程涉及使用Text-to-Speech (TTS)和Dynamic Time Warping (DTW)技术生成真实感的语音轨道，并应用于VidTIMIT和DeepfakeTIMIT数据集以构建新的多模态TIMIT-TTS数据集。该数据集的应用领域包括合成语音检测和多模态深度伪造数据分析，旨在解决当前多媒体取证领域中多模态检测器缺乏的问题。

TIMIT-TTS is a synthetic speech dataset co-developed by Politecnico di Milano and Drexel University, which contains nearly 80,000 speech samples generated via state-of-the-art TTS technologies. This dataset is designed to support research on multimodal synthetic media detection, and can be used either independently or in combination with the DeepfakeTIMIT and VidTIMIT video datasets for multimodal research. The creation of the TIMIT-TTS dataset involves using Text-to-Speech (TTS) and Dynamic Time Warping (DTW) techniques to generate realistic speech tracks, which are then applied to the VidTIMIT and DeepfakeTIMIT datasets to construct the new multimodal TIMIT-TTS dataset. The application scenarios of this dataset cover synthetic speech detection and multimodal deepfake data analysis, with the core goal of addressing the current shortage of multimodal detectors in the field of multimedia forensics.