即時応答型音程トレーナーの為の周波数解析手法について

概要

論文の詳細を見る
It is meaningful to develop a Real-Time Response Pitch Descrimination System by utilizing a computing sound pitch. There are many approaches in computing sound pitch. But, no approach was seen in developing enough response-time and accuracy of frequency discrimination for pitch discrimination training. Therefore new methodology is considered to improve response-time and accuracy of frequency discrimination. In FFT (Fast Fourier Transformation), there exist mathematical relations such that(1) time_range_of_analysis^*frequency_discrimination = 1 and(2) response_time = time_range_of_analysis/2^*computing_time,which shows no improvement in both response-time and accuracy of frequency discrimination. Fortunately, there exist the fact such that many peaks of power spectrum : f[Hz], f^*2[Hz], f^*3[Hz], ･･････ as the results of FFT from a single f[Hz] pitch sound.To improve both response-time and accuracy of frequency discrimination, the following new method is considered in this paper. Step1 : Compute pitch roughly from f[Hz] power spectrum peak. Step2 : Compute pitch accurately from f^*n[Hz].
琉球大学の論文
1994-10-31