Knut Inge
• Member for 4 years, 7 months
• Last seen this week

Signal processing is often at a junction between applied statistics, numerical math, analytic math, physics, electronics and programming. Big data, visualization and machine learning may be ...

A real operation on a complex input can be implemented as two real operations in parallelle, rather than a truely complex operation. An example would be multiplying a complex number with a real number....

A full convolution is Lx + Lh - 1 samples long. This can be truncated to two meaningful subsets, as documented in conv(). filter() is generally an iir filter, and will have an infinite output. Thus it ...

Is it sufficient to identify the «most sinuoidal» of those 3, or would you also want linear projections of those (consistent with a IMU sensor tilted vs the plane of motion)? A simple solution might ...

A file that where the signal path has been kept 96kHz from microphone to loudspeaker has the potential for signal components up to 48kHz. Most physical sound sources, rooms and microphones will roll ...

For a given «visual accuracy», you need to sample the sine at a sufficient number of time-steps per period. At some point the display pixel density will be to low to render a sine accurately. For ...

The image in a digital camera/sensor is represented as a matrix of discrete numbers, as you say. The (spatial) digitization consists of an array of sensels each with some non-zero area that are ...

Crosscorrelation, find the maximum, and use that to offset one file vs the other.

I believe this will do: x=a.^((0:40)+(0:40)’)

I believe that there are diminishing returns beyond radix 4. While arithmetic complexity probably continues to decrease (?) that decrease is too slow to compensate for «other factors». Finding new ...

I think of auto correlation as «self similarity». In practice: line the signal up with a copy of itself. Step the copy one sample at a time. Multiply corresponding pair of samples from the two and ...

It has been said that the DCT reasonably closely matches the KLT for a representative set of images. KLT is essentially the same as PCA, I believe and SVD is only a different way to compute the same? ...

The bandwidth part is somewhat trivial. If you do eg 1000 complex samples/symbols/... per second, that represents twice as much information as 1000 real samples/symbols/... per second. The convenience ...

The transition from text-book math to product code can be a lot simpler with floating point than with fixed point. The latter requires some specialized knowledge that, while not rocket science, may be ...

Generally, it makes sense to ensure that your code is logically correct, that it is numerically well-behaved, intuitive to read and tested. That is hard enough. Only when you observe that some ...

With downsampling you have complete control over the process and it comes down to what compromise of processing complexity, delay, aliasing and loss of passband you can accept. With a lower rate A/D ...

Picture the transmitted signal as a «signature». You want to find some process that maximize the probability of detecting that signature even when there is noise. What do you do to find some signature ...

Could you do something like: (1/2)*unwrap(2*theta)? edit: More concretely, I think about this. I would think that the numerical example and asserts are sufficient to see if it can be adapted to your ...

You are sampling a sine function that should return zero, but because of finite precision argument into that sine function, the error tends to increase towards the right?

The whole point of aliasing and the sampling theoreme is that you cannot (generally) know what an aliased signal was, as you cannot represent infinite bandwidth using finite infornation. If you have ...

In MATLAB, you can do it that way: >> nsamples = 1e9; mu = 11; sigma_squared = 18; x = mu + sqrt(sigma_squared)*randn(nsamples,1); mean(x) ans = 10.9998 var(x) ans = 17.9994 Edit: ...

Unit impulse is a nice pedagogic construct in order to understand how we can measure and execute a LTI system. The stimuli of choice when characterizing real-world near-LTI systems is often not a ...

Yes. The DFT/FFT is a linear function of the audio waveform, and taking the absolute value does not break the proportional relationship. -k

Median filter? Use it to «thin out» bad pixels that appear locally sparse (for some definition of «local»). Then count the remaining bad pixels. Or convolve with a large-ish 2d kernel (eg flat ...

spectrogram(x, N, N/2, N, fs, ‘yaxis’) is my default one liner. To get rid of the non-SI time units, I think you need to do [S, F, T] = spectrogram(x, N, N/2, N, fs); pcolor(S, F, 20*log10(abs(S))); ...

Most Audiophile stuff*) has a credibility on par with homeopathy. It might work - provided that some sales-oriented cable manufacturer single handedly revolutionized science. But in all likelihood it ...

An asynchronous resampler should do the trick. Basically positioning a continuous time windowed sinc at the desired (uniform) output time instants, sampling it by a neighbourhood of input time ...

The high-level picture is that h.265 allows lower bandwidth at the same visual quality (or more visual quality at the same bandwidth). It achieves this by using more advanced techniques that require ...