摘要
研究和改进了一种基于归一化互相关系数的基音检测方法(NCCFPDA)。NCCFPDA算法利用归一化互相关系数、短时归一化能量和自适应门限来区分清/浊音和选取基音周期,计算量小,精度高。在NCCFPDA算法的基础上,采用动态规划的方法对检测出的基音周期进行平滑,并在噪声环境下与自相关法检测基音周期的检测结果进行了比较。实验表明NCCFPDA算法在基音轨迹的平滑和清/浊判别方面都具有令人满意的效果,能实现基音周期的鲁棒性检测。
The NCCFPDA（Normalized Cross-Correlation Function Pitch Detection Algorithm） is studied and improved in this paper. The normalized cross-correlation coefficients, short time normalized energy and adaptive threshold to classify surd/sonant and to select pitch candidates with the hlgh-resolution and efficient implementation are used in NCCFPDA. Based on this algorithm, the dynamic programming to smooth the detected pitch are used and the testing results are compared with the autocorrelation function under the noise environment. Experimental results show that the satisfactory effects for pitch track smooth and surd/sonant decision with this method are gotten and the robust pitch determination is realized.
出处
《电声技术》
2007年第3期54-56,62,共4页
Audio Engineering
关键词
基音周期
归一化互相关系数
动态规划
pitch detection
normalized-cross-correlatlon function
dynamic programming
