当计算快速傅里叶变换时..？
如果我们从具有两个通道的麦克风获取声音，为了计算频谱，我们：
- 计算所有数据输入的快速傅里叶变换？
- 计算左右通道的快速傅立叶变换，然后对两者求平均值？
- 计算左右通道的快速傅里叶变换并获得两者的最大值？
- 以不同的方式计算？
If we get the sound from microphone with two channels, to calculate the frequency spectrum we:
- calculate the fast fourier transform of all the data input?
- calculate the fast fourier transform of each left and right channel and then make a average of both?
- calculate the fast fourier transform of each left and right channel and get the biggest values of both?
- calculate in a different way?
如果你对这篇内容有疑问，欢迎到本站社区发帖提问 参与讨论，获取更多帮助，或者扫码二维码加入 Web 技术交流群。
绑定邮箱获取回复消息
由于您还没有绑定你的真实邮箱，如果其他用户或者作者回复了您的评论，将不能在第一时间通知您！
发布评论
评论（4）
我假设您的双通道麦克风具有物理上独立的输入 - 通常位于演讲厅或剧院的相对两侧。在这种情况下，每个通道的傅立叶变换向您显示该通道输入的物理位置处的频率内容 - 您不能只是对它们进行平均或将它们加在一起并获得任何有意义的结果。
如果您试图获得通道输入之间中点的频谱，则平均可能有意义，但其他任何方法都需要一些大假设（反射不是因素等）和插值。
对这些数据的任何有意义的分析都必须考虑声音采样空间的物理几何形状。明确您要寻找的内容。
哦，并将您的帖子标记为作业。
I assume your two-channel microphone has inputs that are physically separate - typically, opposite sides of a lecture hall or theater. In this case, the Fourier transform of each channel shows you the frequency content at the physical location of that channel's input - you can't just average them or add them together and get any meaningful result.
If you're trying to get the spectrum at the midpoint between the channel inputs, averaging might make sense, but anything else would require some big assumptions (reflection not a factor, etc) and interpolation.
Any meaningful analysis of this data is going to have to take into account the physical geometry of the space where the sound is sampled. Clarify what you're looking for.
Oh, and tag your posts as homework.
FFT 是一种线性变换。因此，获得平均左右输入的 FFT 完全等同于分别获得左通道和右通道的两个 FFT 的平均值。
但通常将它们分开是有意义的 - 或者仅使用其中一种渠道。
FFT is a linear transform. Thus, getting FFT of averaged left-right input is completely equivalent to getting average of two FFTs, for left and right channel respectively.
But usually it makes sense to keep them separate - or use only one of the channels.
在不知道您的设置是什么以及此处理的目的的情况下，我们只能猜测一个好的答案。我假设你的意思是你在不同的位置有两个麦克风。
如果先将两个通道相加，或者对每个通道进行傅立叶变换然后相加，则会得到不好的结果。原因是干扰 - 在某些频率下，源将是来自一个麦克风的整数波长，但与另一个麦克风的波长的整数倍半。这是初学者在录音时常见的问题。
也许您想要测量两个麦克风听到的源信号，同时忽略每个麦克风本地的随机噪声和无关源。在这种情况下，对每个通道进行傅立叶运算，计算每个频率的功率（幅度的平方），取其对数，然后对两个通道进行平均。这避免了干扰问题，并且统计上减少了随机噪声（尽管不是很多）。
Without knowing just what your setup is, and the purpose of this processing, we can only guess at a good answer. I'll assume you mean that you have two microphones in different locations.
If you add the two channels first, or if you Fourier transform each and then add, you'll get bad results. The reason is interference - at some frequencies the source will be an integer number of wavelengths from one microphone but an integer and a half from the other. This is a common beginner's problem in audio recording.
Perhaps what you want to it measure the signal of a source heard by both microphones, while ignoring random noise and extraneous sources local to each microphone. In that case, do a Fourier on each channel, compute the power (the squared magnitude) at each frequency, take the logarithm of that, and then average the two channels. This avoid problems with interference, and statistically reduces the random noise (though not a whole lot).
要结合这两个 FT，您必须首先了解它们 TOA（到达时间）的差异并加以考虑。查找波束成形。如果我们对应用程序有一些了解将会有所帮助。正如其他答案中所述，当然您可以对两者进行平均并了解某个区域的频谱，但目的是什么？请提供有关目的的更多详细信息，然后我们可以告诉您您所提议的是否合适。
To combine these two FT's you must know the difference in their TOA (time of arrival) and account for it first. Look up beam forming. It would help if we had some idea of the application. As has been said in the other answers, sure you can average the two and get some idea of the frequency spectrum across an area, but to what end? More details please as to the purpose, then we can tell you whether or not what you're proposing is a decent fit.