LRPD: Large Replay Parallel Dataset

LRPD数据集由ID R&D Inc.创建，旨在推动语音反欺骗技术的发展，特别是针对重放攻击的检测。该数据集包含超过100万条语音记录，由19种不同设备在17种环境中采集。数据集内容丰富，包括多种语言和方言，以及不同设备和环境下的录音。创建过程中，采用了多设备并行录音技术，确保数据的高质量和多样性。LRPD数据集主要应用于自动语音识别系统的安全性增强，解决语音重放攻击的检测问题。

The LRPD dataset was created by ID R&D Inc. with the goal of advancing speech anti-spoofing technologies, specifically for replay attack detection. It contains over one million speech recordings collected by 19 different devices across 17 environments. The dataset boasts rich and diverse content, encompassing multiple languages and dialects, along with recordings obtained under various device and environmental conditions. During its development, multi-device parallel recording technology was utilized to ensure the high quality and diversity of the dataset. The LRPD dataset is primarily applied to enhance the security of automatic speech recognition (ASR) systems by tackling the issue of speech replay attack detection.

搜集汇总

背景与挑战

背景概述

LRPD数据集是一个大规模语音数据集，专为语音反欺骗技术设计，特别针对重放攻击检测。它包含超过100万条记录，由19种设备在17种环境中采集，涵盖多种语言和方言，采用多设备并行录音技术确保数据多样性和高质量。该数据集主要用于增强自动语音识别系统的安全性，解决重放攻击问题。