Towards Improving Speaker Distance Estimation through Generative Impulse Response Augmentation
通过生成性冲击响应增强改进说话距离估计
机构 * Amazon(亚马逊) ; Wentworth Institute of Technology(韦尔特沃斯技术学院)
AI总结 本文通过生成冲击响应数据增强提升说话距离估计模型性能,采用FastRIR生成冲击响应并优化超参数,实验显示在中长距离显著提高估计精度。
Comments Accepted to Generative Data Augmentation for Real-World Signal Processing Applications (GenDA 2025). An ICASSP 2025 Satellite Workshop and IEEE Data Science and Learning Workshop: Room Acoustics and Speaker Distance Estimation Challenge
Journal ref Generative Data Augmentation for Real-World Signal Processing Applications (GenDA 2025). An ICASSP 2025 Satellite Workshop and IEEE Data Science and Learning Workshop