I am quite new to signal processing so forgive me if I rant on a bit. I have download and installed FFTW for windows. The documentation is ok but I still have queries.
My overall aim is to capture raw audio data sampled at 44100 samps/sec from the sound card on the computer (this task is already implemented using libraries and my code), and then perform the DFT on blocks of this audio data.
I am only interested in finding a range of frequency components in the audio and I will not be performing any inverse DFT. In this case, is a real to real transformation all that is necessary, hence the fftw_plan_r2r_1d() function?
My blocks of data to be transformed are 11025 samples long. My function is called as shown below. This will result in a spectrum array of 11025 bins. How do I know the maximum frequency component in the result?
I believe that the bin spacing is Fs/n , 44100/11025, so 4. Does it mean that I will have a frequency spectrum in the array from 0 Hz all the way up to 44100Hz in steps of 4, or up to half the nyquist frequency 22200?
This would be a problem for me as I only wish to search for frequencies from 60Hz up to 3000Hz. Is there some way to limit the transform range?
I don't see any arguments for the function, or maybe there is another way?
Many thanks in advance for any help with this.
p = fftw_plan_r2r_1d(11025, audioData, spectrum, FFTW_REDFT00, FFTW_ESTIMATE);
To answer some of your individual questions from the above:
you need a real-to-complex transform, not real-to-real
you will calculate the magnitude of the complex output bins at the frequencies of interest (magnitude = sqrt(re*re + im*im))
the frequency resolution is indeed Fs / N = 44100 / 11025 = 4 Hz, i.e. the width of each output bin is 4 Hz
for a real-to-complex transform you get N/2 + 1 output bins giving you frequencies from 0 to Fs / 2
you just ignore frequencies in which you are not interested - the FFT is very efficient so you can afford to "waste" unwanted output bins (unless you are only interested in a relatively small number of output frequencies)
Additional notes:
plan creation does not actually perform an FFT - typically you create a plan once and then use it many times (by calling fftw_execute)
for performance you probably want to use the single precision calls (e.g. fftwf_execute rather than fftw_execute, and similarly for plan creation etc)
Some useful related questions/answers on StackOverflow:
How do I obtain the frequencies of each value in an FFT?
How to get frequency from fft result?
How to generate the audio spectrum using fft in C++?
There are many more similar questions and answers which you might also want to read - search for the fft and fftw tags.
Also note that dsp.stackexchange.com is the preferred site for site for questions on DSP theory rather than actual specific programming problems.
I want to make a frequency shift on a .wav file.
The problem I have is that the FFT uses complex numbers, and the .wav file has integer values. So I want to make a frequency shift and that means that I have to make a direct transform and an inverse transform , the problem is that the inverse transform doesn't give me integer values (it gives me complex values), but I need integer values for the samples of the .wav file.
How do I interpret the values of the inverse transform ?
I want to make a frequency shift on a .wav file.
So you've got an audio, which means a real-valued, signal.
The spectrum of a real valued signal has symmetry to f=0, i.e. it's Fourier Transform has hermitian symmetry.
If you now shift that input spectrum (blue), the result (red) loses symmetry, i.e. the resulting signal is no longer real:
Notice how things are, through aliasing, circular, so what you "shift out" of the Nyquist range will appear at the opposite end. In my example, this means that you get unexpected high frequency components!
The problem I have is that the FFT uses complex numbers, and the .wav file has integer values. So I want to make a frequency shift and that means that I have to make a direct transform and an inverse transform , the problem is that the inverse transform doesn't give me integer values (it gives me complex values), but I need integer values for the samples of the .wav file.
Indeed! And that's because the result of your shift really isn't a real signal anymore.
What you can do, however, is:
shift (either in time or frequency domain – honestly, doing it in time domain would be easier! Just multiply the nth sample with exp(2j π f_shift/f_sample n ).)
apply a complex bandpass filter that removes everything outside the frequency [0; f_sample / 2 - shift ]. This gives you what is called an analytical signal (i.e. only positive frequencies), which still isn't real-valued, because it's not symmetric.
Throwing away the imaginary part now doesn't alter the information of your signal - it just halves the energy and gives you a symmetric spectrum, and something you can write to a .wav file.
Now, this whole "doing it in frequency domain through the FFT" is an approach that people in the Software Defined Radio world are pretty used to – they deal with complex baseband signals all the time.
How do I interpret the values of the inverse transform ?
As the complex signal that they are. Ignoring the imaginary part as suggested in the comments will lead to the energy contained in the negative frequencies being mirrored onto you positive frequencies (and the other way around), and is most likely not what you want – unless either:
you've made sure that prior to this "symmetricalization", the energy on either side of f=0 was 0, so that nothing bad happens, pr
you've made sure that you selectively shifted the negative and positive frequencies so, that symmetry was retained. Notice that this is not a "simple" shift of the whole frequency domain, but two selective shifts; the selection of these shifting regions has a shape, which boils down to using a window. If you just "select" or "not select" each bin for shifting, you're effectively applying a rectangular window – with all the Gibb's Phenomenom you can incur with that.
I don't understand how someone could come up with a simple 3x3 matrix called kernel, so when applied to the image, it would produce some awesome effect. Examples: http://en.wikipedia.org/wiki/Kernel_(image_processing) . Why does it work? How did people come up with those kernels (trial and error?)? Is it possible to prove it will always work for all images?
I don't understand how someone could come up with a simple 3x3 matrix called kernel, so when applied to the image, it would produce some awesome effect. Examples: http://en.wikipedia.org/wiki/Kernel_(image_processing).
If you want to dig into the history, you'll need to check some other terms. In older textbooks on image processing, what we think of as kernels today are more likely to be called "operators." Another key term is convolution. Both these terms hint at the mathematical basis of kernels.
http://en.wikipedia.org/wiki/Convolution
You can read about mathematical convolution in the textbook Computer Vision by Ballard and Brown. The book dates back to the early 80s, but it's still quite useful, and you can read it for free online:
http://homepages.inf.ed.ac.uk/rbf/BOOKS/BANDB/toc.htm
From the table of contents to the Ballard and Brown book you'll find a link to a PDF for section 2.2.4 Spatial Properties.
http://homepages.inf.ed.ac.uk/rbf/BOOKS/BANDB/LIB/bandb2_2.pdf
In the PDF, scroll down to the section "The Convolution Theorem." This provides the mathematical background for convolution. It's a relatively short step from thinking about convolution expressed as functions and integrals to the application of the same principles to the discrete world of grayscale (or color) data in 2D images.
You will notice that a number of kernels/operators are associated with names: Sobel, Prewitt, Laplacian, Gaussian, and so on. These names help suggest that there's a history--really quite a long history--of mathematical development and image processing research that has lead to the large number of kernels in common use today.
Gauss and Laplace lived long before us, but their mathematical work has trickled down into forms we can use in image processing. They didn't work on kernels for image processing, but mathematical techniques they developed are directly applicable and commonly used in image processing. Other kernels were developed specifically for processing images.
The Prewitt operator (kernel), which is quite similar to the Sobel operator, was published in 1970, if Wikipedia is correct.
http://en.wikipedia.org/wiki/Prewitt_operator
Why does it work?
Read about the mathematical theory of convolution to understand how one function can be "passed over" or "dragged" across another. That can explain the theoretical basis.
Then there's the question of why individual kernels work. In you look at the edge transition from dark to light in an image, and if you plot the pixel brightness on a 2D scatterplot, you'll notice that the values in the Y-axis increase rapidly about the edge transition in the image. That edge transition is a slope. A slope can be found using the first derivative. Tada! A kernel that approximates a first derivative operator will find edges.
If you know there's such a thing in optics as Gaussian blur, then you might wonder how it could be applied to a 2D image. Thus the derivation of the Gaussian kernel.
The Laplacian, for instance, is an operator that, according to the first sentence from the Wikipedia entry, "is a differential operator given by the divergence of the gradient of a function on Euclidean space."
http://en.wikipedia.org/wiki/Laplacian
Hoo boy. It's quite a leap from that definition to a kernel. The following page does a fine job of explaining the relationship between derivatives and kernels, and it's a quick read:
http://www.aishack.in/2011/04/the-sobel-and-laplacian-edge-detectors/
You'll also see that one form of the Laplacian kernel is simply named the "edge-finding" kernel in the Wikipedia entry you cited.
There is more than one edge-finding kernel, and each has its place. The Laplacian, Sobel, Prewitt, Kirsch, and Roberts kernels all yield different results, and are suited for different purposes.
How did people come up with those kernels (trial and error?)?
Kernels were developed by different people following a variety of research paths.
Some kernels (to my memory) were developed specifically to model the process of "early vision." Early vision isn't what happens only to early humans, or only for people who rise at 4 a.m., but instead refers to the low-level processes of biological vision: sensing of basic color, intensity, edges, and that sort of thing. At the very low level, edge detection in biological vision can be modeled with kernels.
Other kernels, such as the Laplacian and Gaussian, are approximations of mathematical functions. With a little effort you can derive the kernels yourself.
Image editing and image processing software packages will often allow you to define your own kernel. For example, if you want to identify a shape in an image small enough to be defined by a few connected pixels, then you can define a kernel that matches the shape of the image feature you want to detect. Using custom kernels to detect objects is too crude to work in most real-world applications, but sometimes there are reasons to create a special kernel for a very specific purpose, and sometimes a little trial and error is necessary to find a good kernel.
As user templatetypedef pointed out, you can think of kernels intuitively, and in a fairly short time develop a feel for what each would do.
Is it possible to prove it will always work for all images?
Functionally, you can throw a 3x3, 5x5, or NxN kernel at an image of the appropriate size and it'll "work" in the sense that the operation will be performed and there will be some result. But then the ability to compute a result whether it's useful or not isn't a great definition of "works."
One information definition of whether a kernel "works" is whether convolving an image with that kernel produces a result that you find useful. If you're manipulating images in Photoshop or GIMP, and if you find that a particular enhancement kernel doesn't yield quite what you want, then you might say that kernel doesn't work in the context of your particular image and the end result you want. In image processing for computer vision there's a similar problem: we must pick one or more kernels and other (often non-kernel based) algorithms that will operate in sequence to do something useful such as identify faces, measures the velocity of cars, or guide robots in assembly tasks.
Homework
If you want to understand how you can translate a mathematical concept into a kernel, it helps to derive a kernel by yourself. Even if you know what the end result of the derivation should be, to grok the notion of kernels and convolution it helps to derive a kernel from a mathematical function by yourself, on paper, and (preferably) from memory.
Try deriving the 3x3 Gaussian kernel from the mathematical function.
http://en.wikipedia.org/wiki/Gaussian_function
Deriving the kernel yourself, or at least finding an online tutorial and reading closely, will be quite revealing. If you'd rather not do the work, then you may not appreciate the way that some mathematical expression "translates" to a bunch of numbers in a 3x3 matrix. But that's okay! If you get the general sense of a common kernel is useful, and if you observe how two similar kernels produce slightly different results, then you'll develop a good feel for them.
Intuitively, a convolution of an image I with a kernel K produces a new image that's formed by computing a weighted sum, for each pixel, of all the nearby pixels weighted by the weights in K. Even if you didn't know what a convolution was, this idea still seems pretty reasonable. You can use it to do a blur effect (by using a Gaussian weighting of nearby pixels) or to sharpen edges (by subtracting each pixel from its neighbors and putting no weight anywhere else.) In fact, if you knew you needed to do all these operations, it would make sense to try to write a function that given I and K did the weighted sum of nearby pixels, and to try to optimize that function as aggressively as possible (since you'd probably use it a lot).
To get from there to the idea of a convolution, you'd probably need to have a background in Fourier transforms and Fourier series. Convolutions are a totally natural idea in that domain - if you compute the Fourier transformation of two images and multiply the transforms together, you end up computing the transform of the convolution. Mathematicians had worked that out a while back, probably by answering the very natural question "what function has a Fourier transform defined by the product of two other Fourier transforms?," and from there it was just a matter of time before the connection was found. Since Fourier transforms are already used extensively in computing (for example, in signal processing in networks), my guess is that someone with a background in Fourier series noticed that they needed to apply a kernel K to an image I, then recognized that this is way easier and more computationally efficient when done in frequency space.
I honestly have no idea what the real history is, but this is a pretty plausible explanation.
Hope this helps!
There is a good deal of mathematical theory about convolutions, but the kernel examples you link to are simple to explain intuitively:
[ 0 0 0]
[ 0 1 0]
[ 0 0 0]
This one says to take the original pixel and nothing else, so it yields just the original image.
[-1 -1 -1]
[-1 8 -1]
[-1 -1 -1]
This one says to subtract the eight neighbors from eight times the original pixel. First consider what happens in a smooth part of the image, where there is solid, unchanging color. Eight times the original pixel equals the sum of eight identical neighbors, so the difference is zero. Thus, smooth parts of the image become black. However, parts of the images where there are changes do not become black. Thus, this kernel highlights changes, so it highlights places where one shape ends and another begins: the edges of objects in the image.
[ 0 1 0]
[ 1 -4 1]
[ 0 1 0]
This is similar to the one above, but it is tuned differently.
[ 0 -1 0]
[-1 5 -1]
[0 -1 0]
Observe that this is just the negation of the edge detector above plus the first filter we saw, the one for the original image. So this kernel both highlights edges and adds that to the original image. The result is the original image with more visible edges: a sharpening effect.
[ 1 2 1]
[ 2 4 2]
[ 1 2 1]
[ 1 1 1]
[ 1 1 1]
[ 1 1 1]
Both of these blend the original pixel with its neighbors. So they blur the image a little.
There are two ways of thinking about (or encoding) an image: the spatial domain and the frequency domain. A spatial representation is based on pixels, so it's more familiar and easier to obtain. Both the image and the kernel are expressed in the spatial domain.
To get to the frequency domain, you need to use a Fourier or related transform, which is computationally expensive. Once you're there, though, many interesting manipulations are simpler. To blur an image, you can just chop off some high-frequency parts — like cropping the image in the spatial domain. Sharpening is the opposite, akin to increasing the contrast of high-frequency information.
Most of the information of an image is in the high frequencies, which represent detail. Most interesting detail information is at a small, local scale. You can do a lot by looking at neighboring pixels. Blurring is basically taking a weighted average of neighboring pixels. Sharpening consists of looking at the difference between a pixel and its neighbors and enhancing the contrast.
A kernel is usually produced by taking a frequency-domain transformation, then keeping only the high-frequency part and expressing it in the spatial domain. This can only be done for certain transformation algorithms. You can compute the ideal kernel for blurring, sharpening, selecting certain kinds of lines, etc., and it will work intuitively but otherwise seems like magic because we don't really have a "pixel arithmetic."
Once you have a kernel, of course, there's no need to get into the frequency domain at all. That hard work is finished, conceptually and computationally. Convolution is pretty friendly to all involved, and you can seldom simplify any further. Of course, smaller kernels are friendlier. Sometimes a large kernel can be expressed as a convolution of small sub-kernels, which is a kind of factoring in both the math and software senses.
The mathematical process is pretty straightforward and has been studied since long before there were computers. Most common manipulations can be done mechanically on an optical bench using 18th century equipment.
I think the best way to explain them is to start in 1d and discuss the z-transform and its inverse. That switches from the time domain to the frequency domain — from describing a wave as a timed sequence of samples to describing it as the amplitude of each frequency that contributes to it. The two representations contain the same amount of information, they just express it differently.
Now suppose you had a wave described in the frequency domain and you wanted to apply a filter to it. You might want to remove high frequencies. That would be a blur. You might want to remove low frequencies. That would be a sharpen or, in extremis, an edge detect.
You could do that by just forcing the frequencies you don't want to 0 — e.g. by multiplying the entire range by a particular mask, where 1 is a frequency you want to keep and 0 is a frequency you want to eliminate.
But what if you want to do that in the time domain? You could transfer to the frequency domain, apply the mask, then transform back. But that's a lot of work. So what you do (approximately) is transform the mask from the frequency domain to the time domain. You can then apply it in the time domain.
Following the maths involved for transforming back and forth, in theory to apply that you'd have to make each output sample a weighted sum of every single input sample. In the real world you make a trade-off. You use the sum of, say, 9 samples. That gives you a smaller latency and less processing cost than using, say, 99 samples. But it also gives you a less accurate filter.
A graphics kernel is the 2d analogue of that line of thought. They tend to be small because processing cost grows with the square of the edge length so it gets expensive very quickly. But you can approximate any sort of frequency domain limiting filter.
I have implemented a lookup table to compute sine/cosine values in my system. I now need inverse trigonometric functions (arcsin/arccos).
My application is running on an embedded device on which I can't add a second lookup table for arcsin as I am limited in program memory. So the solution I had in mind was to browse over the sine lookup table to retrieve the corresponding index.
I am wondering if this solution will be more efficient than using the standard implementation coming from the math standard library.
Has someone already experimented on this?
The current implementation of the LUT is an array of the sine values from 0 to PI/2. The value stored in the table are multiplied by 4096 to stay with integer values with enough precision for my application. The lookup table as a resolution of 1/4096 which give us an array of 6434 values.
Then I have two funcitons sine & cosine that takes an angle in radian multiplied by 4096 as argument. Those functions convert the given angle to the corresponding angle in the first quadrant and read the corresponding value in the table.
My application runs on dsPIC33F at 40 MIPS an I use the C30 compiling suite.
It's pretty hard to say anything with certainty since you have not told us about the hardware, the compiler or your code. However, a priori, I'd expect the standard library from your compiler to be more efficient than your code.
It is perhaps unfortunate that you have to use the C30 compiler which does not support C++, otherwise I'd point you to Optimizing Math-Intensive Applications with Fixed-Point Arithmetic and its associated library.
However the general principles of the CORDIC algorithm apply, and the memory footprint will be far smaller than your current implementation. The article explains the generation of arctan() and the arccos() and arcsin() can be calculated from that as described here.
Of course that suggests also that you will need square-root and division also. These may be expensive though PIC24/dsPIC have hardware integer division. The article on math acceleration deals with square-root also. It is likely that your look-up table approach will be faster for the direct look-up, but perhaps not for the reverse search, but the approaches explained in this article are more general and more precise (the library uses 64bit integers as 36.28 bit fixed point, you might get away with less precision and range in your application), and certainly faster than a standard library implementation using software-floating-point.
You can use a "halfway" approach, combining a coarse-grained lookup table to save memory, and a numeric approximation for the intermediate values (e.g. Maclaurin Series, which will be more accurate than linear interpolation.)
Some examples here.
This question also has some related links.
A binary search of 6434 will take ~12 lookups to find the value, followed by an interpolation if more accuracy is needed. Due to the nature if the sin curve, you will get much more accuracy at one end than the other. If you can spare the memory, making your own inverse table evenly spaced on the inputs is likely a better bet for speed and accuracy.
In terms of comparison to the built-in version, you'll have to test that. When you do, pay attention to how much the size of your image increases. The stdin implementations can be pretty hefty in some systems.
Let us say that I have a WAV file. In this file, is a series of sine tones at precise 1 second intervals. I want to use the FFTW library to extract these tones in sequence. Is this particularly hard to do? How would I go about this?
Also, what is the best way to write tones of this kind into a WAV file? I assume I would only need a simple audio library for the output.
My language of choice is C
To get the power spectrum of a section of your file:
collect N samples, where N is a power of 2 - if your sample rate is 44.1 kHz for example and you want to sample approx every second then go for say N = 32768 samples.
apply a suitable window function to the samples, e.g. Hanning
pass the windowed samples to an FFT routine - ideally you want a real-to-complex FFT but if all you have a is complex-to-complex FFT then pass 0 for all the imaginary input parts
calculate the squared magnitude of your FFT output bins (re * re + im * im)
(optional) calculate 10 * log10 of each magnitude squared output bin to get a magnitude value in dB
Now that you have your power spectrum you just need to identify the peak(s), which should be pretty straightforward if you have a reasonable S/N ratio. Note that frequency resolution improves with larger N. For the above example of 44.1 kHz sample rate and N = 32768 the frequency resolution of each bin is 44100 / 32768 = 1.35 Hz.
You are basically interested in estimating a Spectrum -assuming you've already gone past the stage of reading the WAV and converting it into a discrete time signal.
Among the various methods, the most basic is the Periodogram, which amounts to taking a windowed Discrete Fourier Transform (with a FFT) and keeping its squared magnitude. This correspond to Paul's answer. You need a window which spans over several periods of the lowest frequency you want to detect. Example: if your sinusoids can be as low as 10 Hz (period = 100ms), you should take a window of 200ms o 300ms or so (or more). However, the periodogram has some disadvantages, though it's simple to compute and it's more than enough if high precision is not required:
The raw periodogram is not a good
spectral estimate because of spectral
bias and the fact that the variance
at a given frequency does not decrease
as the number of samples used in the
computation increases.
The periodogram can perform better by averaging several windows, with a judious choosing of the widths (Bartlet method). And there are many other methods for estimating the spectrum (AR modelling).
Actually, you are not exactly interested in estimating a full spectrum, but only the location of a single frequency. This can be done seeking a peak of an estimated spectrum (done as explained), but also by more specific and powerful (and complicated) methods (Pisarenko, MUSIC algorithm). They would probably be overkill in your case.
WAV files contain linear pulse code modulated (LPCM) data. That just means that it is a sequence of amplitude values at a fixed sample rate. A RIFF header is contained at the beginning of the file to convey information like sampling rate and bits per sample (e.g. 8 kHz signed 16-bit).
The format is very simple and you could easily roll your own. However, there are several libraries available to speed the process such as libsndfile. Simple Direct-media Layer (SDL)/SDL_mixer and PortAudio are two nice libraries for playback.
As for feeding the data into FFTW, you would need to buffer 1 second chunks (determine size by the sample rate and bits per sample). Then convert all of the samples to IEEE floating-point (i.e. float or double depending on the FFTW configuration--libsndfile can do this for you). Next create another array to hold the frequency domain output. Finally, create and execute an FFTW plan by passing both buffers to fftw_plan_dft_r2c_1d and calling fftw_execute with the returned fftw_plan handle.