240小时印度印地语语音数据_朗读（手机）【数据堂】

印度印地语语音数据_朗读（手机），基于给定的脚本朗读并模拟录制，覆盖经济，娱乐，新闻，口语等多个领域，内容丰富。此数据集标注了文本内容、时间戳等多种属性，由401名来自不同地域和文化背景的印度人参与录制，口音正宗，录音涵盖安静和噪音的不同环境，更贴合语音识别实际应用场，文本经过人工校对，准确率高，为语音识别相关研究及应用提供了丰富的资源，经多家AI公司验证：有助于模型面对真实世界的多样性时能够表现出色。我们严格遵循数据保护法规和隐私规定，确保数据采集、存储和使用的过程中维护用户的隐私和合法权益，所有数据均遵循GDPR, CCPA, PIPL

Indian Hindi Read Speech Dataset (Recorded via Mobile Phones): This dataset is developed based on scripted reading and simulated recording sessions, covering diverse domains such as economy, entertainment, news, and colloquial speech, thus possessing rich and varied content. Multiple attributes including text transcripts and timestamps are annotated for this dataset. A total of 401 Indian participants from different regions and cultural backgrounds participated in the recording, producing authentic regional accents. The recordings are collected in both quiet and noisy environments, which aligns closely with real-world application scenarios of speech recognition systems. All text transcripts have undergone manual proofreading to ensure high accuracy, providing valuable resources for research and applications related to speech recognition. It has been validated by multiple AI enterprises that this dataset enables speech recognition models to achieve robust performance when confronted with real-world linguistic and environmental diversities. We strictly adhere to data protection regulations and privacy provisions, ensuring the protection of participants' privacy and legitimate rights and interests throughout the entire process of data collection, storage, and utilization. All datasets comply with GDPR, CCPA and PIPL.

搜集汇总

背景与挑战

背景概述

该数据集包含240小时由401名印度人使用手机录制的印地语朗读语音，覆盖财经、娱乐、新闻等多个领域，并标注了文本内容、时间戳等属性。录音环境包括安静和噪音条件，格式为16kHz WAV，句准确率达95%，旨在为语音识别研究和应用提供多样化的真实世界资源。