2021 SLT CSRC

儿童语音识别对发展智能至关重要 儿童与机器交互和计算机辅助等技术 语言学习。但儿童的声学和语言模式 言语与成人言语明显不同。此外 如今，儿童语音数据集仍然有限，这可能 阻碍了儿童语音识别研究的发展。 关于这一点，我们推出了 儿童语音识别挑战赛 ，作为旗舰 IEEE SLT 2021研讨会的卫星活动，其中 将为注册团队发布约400小时的数据，并设置两个 挑战赛道，旨在提高儿童的语音识别能力 研究。希望挑战赛也能为以下方面提供良好的试验台： 相关技术，如领域适应、迁移学习等 上。Child speech recognition is critically important for the development of intelligence, as well as technologies including human-machine interaction and computer-assisted language learning. However, the acoustic and linguistic patterns of children’s speech differ significantly from those of adult speech. Furthermore, the current availability of child speech datasets remains limited, which may hinder the progress of child speech recognition research. To address this issue, we launched the Child Speech Recognition Challenge as a satellite event of the flagship IEEE SLT 2021 workshop. Approximately 400 hours of data will be released for registered teams, and two challenge tracks will be established, aimed at advancing child speech recognition research. We hope that this challenge will also serve as an excellent testbed for related technologies such as domain adaptation, transfer learning, and other similar techniques.

搜集汇总

背景与挑战

背景概述

2021 SLT CSRC是一个专注于儿童语音识别的数据集，包含约400小时的儿童语音数据，旨在推动儿童语音识别技术的研究和发展。该数据集由清华大学、西北工业大学和标贝科技联合发布，并作为IEEE SLT 2021研讨会的卫星活动推出。