数据堂—377小时河南方言自然对话手机采集语音数据

377小时河南方言自然对话手机采集语音数据由762名河南本土人进行录制，录音人覆盖多个年龄段，男女比例均衡，377小时河南方言自然对话手机采集语音数据不指定话题，无预设文本，由录音人两两组合自由交谈，可用于语音识别声学、声纹识别模型训练或算法研究。

A 377-hour mobile-collected speech dataset of natural conversational Henan dialect was recorded by 762 native Henan speakers. The participating speakers span multiple age groups with a balanced gender ratio. The conversations are unscripted and topic-free: no pre-defined texts are provided, and the speakers are paired freely to engage in spontaneous talks. This dataset can be applied to acoustic speech recognition model training, speaker verification model training, and related algorithm research.

搜集汇总

背景与挑战

背景概述

该数据集包含377小时河南方言自然对话语音数据，由762名本地人通过手机录制，内容为无预设话题的自由对话，采用16kHz/16bit的wav格式，主要用于方言语音识别模型测试。