让没有大样本来KissFFT [英] Giving large no of samples to KissFFT

问题描述

我想找到的持续时间10秒钟的采样率44100赫兹的音频信号的4096点DFT。因此有441000输入采样。但KissFFT只需要达到4096作为输入的大小。如何去寻找这样一个大信号的FFT？

I wanted to find the 4096 point DFT of an audio signal of duration 10 seconds with sampling rate 44100 Hz. Hence there are 441000 input samples. But KissFFT takes only up to 4096 as input size. How to go about finding FFT of such a large signal?

推荐答案

最真实世界的音频信号（语音，音乐等）的功率谱是时变的，所以通常你计算了一系列短期的FFT使用重叠窗口，来产生功率谱序列，又名一个的频谱的

The power spectrum of most real-world audio signals (speech, music, etc) is time-varying, so typically you calculate a series of short-term FFTs using overlapping windows, to produce a sequence of power spectra, aka a spectrogram.

我建议开始用50％的重叠，所以第一FFT将样品0..4095，第二个针对2048..6143等

I suggest starting with a 50% overlap, so the first FFT would be for samples 0..4095, the second for 2048..6143, etc.

这篇关于让没有大样本来KissFFT的文章就介绍到这了，希望我们推荐的答案对大家有所帮助，也希望大家多多支持IT屋！