提出一种抗噪混响的音高估计方法，提升语音处理在恶劣环境下的准确率。
Harmonic Summation-Based Robust Pitch Estimation in Noisy and Reverberant Environments
- 用归一化幅度差函数生成概率音高状态，结合多周期与邻帧聚合增强鲁棒性
- 在不同信噪比下均低于现有方法的总音高误差和语音判定误差
- 适合需要高精度音高的语音识别、语音合成等应用
准确的音高估计对众多语音处理任务至关重要，但在高失真环境下仍具挑战。本文提出一种鲁棒音高估计方法，在复杂噪声环境中仍能提供可靠估计。该方法计算归一化平均幅度差函数（NAMDF），将其转换为似然函数，并为每个样本偏移量生成帧级的概率音高状态。为增强抗噪能力，我们在整数倍音高周期及相邻帧间聚合似然值。此外，通过在维特比算法中引入简单有效的连续性约束，优化多个候选音高间的选择。实验结果表明，该方法在不同信噪比（SNR）条件下均显著降低总音高误差（GPE）和语音判定误差（VDE），在噪声和混响环境下均优于现有方法。
原文摘要 · Abstract (English)
Accurate pitch estimation is essential for numerous speech processing applications, yet it remains challenging in high-distortion environments. This paper proposes a robust pitch estimation method that delivers robust pitch estimates in challenging noise environments. Our approach computes the Normalized Average Magnitude Difference Function (NAMDF), transforms it into a likelihood function, and generates probabilistic pitch states for frames at each sample shift. To enhance noise robustness, we aggregate likelihood values across integer multiples of the pitch period and neighboring frames. Furthermore, we introduce a simple yet effective continuity constraint in the Viterbi algorithm to refine pitch selection among multiple candidates. Experimental results show that our method consistently achieves lower Gross Pitch Error (GPE) and Voicing Decision Error (VDE) across various SNR levels, outperforming existing methods in both noisy and reverberant conditions.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。