7 votes

Can deep neural networks achieve real-time video analysis?

Nvidia seems to have published some white papers comparing DNN inference performance between high-powered CPUs and (of course) Nvidia GPUs. (one example) Ballpark seems to be that some systems can ...
hotpaw2's user avatar
  • 35.3k
6 votes

Is there a penis-detection demo similar to face-detection?

Is there a script / tutorial / demo for penis detection? [...] Fairly serious quesion, future of internet memes is at stake Yes, there is. Common Pattern Recognition techniques will be able to spot ...
A_A's user avatar
  • 10.7k
6 votes

Neural Networks and Complex Valued Inputs

The power of complex representations remains an open topic to me. I still do strive the understand Fourier transformations. An underlying question is, to me: why would complex transformations be ...
Laurent Duval's user avatar
6 votes
Accepted

Downsampling audio for use in Machine Learning

Is this necessary if I only intend to use the data in the Neural Network toolkit provided by the repo I linked? Yes. Whether or not you are downsampling (instead of just decimating) has nothing to ...
A_A's user avatar
  • 10.7k
5 votes
Accepted

Spectrograms for neural nets

The "dimensions" of the spectrogram are not chosen based on where will the spectrogram be fed to but rather depend on your application. Therefore, it is key to understand the spectrogram itself first, ...
A_A's user avatar
  • 10.7k
3 votes
Accepted

Pattern recognition in time series 4x3000 vector

Assuming your time series are the same length, take your data and produce spectrograms, $S_k$, for each row of the $4 \times 3000$ data matrix $D$ that you have. Since these individual time series ...
spektr's user avatar
  • 263
3 votes

Convolution and Cross Correlation on 2D Image

If as you said you understand well the 1-D convolution/cross-correlation functioning (the Wikipedia first graph explains it in a clear way), the 2-D version is very similar! This website explains 2-D ...
Louis Lac's user avatar
  • 378
3 votes

What is the best input for de-noising autoencoder for sound data?

As you mentioned MFCC features are one of the best features to represent audio as it captures both the time and frequency variations in the audio clip.You can get more details about MFCCS features in ...
Navaratnarajah Suman's user avatar
3 votes
Accepted

Deriving the Langrangian interpolation polynomials in Cook-Toom convolutions

If you represent a second-order polynomial $s(x)$ with Lagrange polynomials $L_i(x)$ and interpolation points $\beta_i$, $i=0,1,2$, such that $$s(x)=s(\beta_0)L_0(x)+s(\beta_1)L_1(x)+s(\beta_2)L_2(x)\...
Matt L.'s user avatar
  • 90k
3 votes

Upsampling vs downsampling. Which to use when?

You should use the one you need for your problem, when you know which components of your signal are of interest to you. Let's say you have in your electronic editing an ADC digitizing 40M samples per ...
Nathan Huchon's user avatar
3 votes

a neural network approach for FIR filter

Yes. The FIR filter model you're used to is a series of Neurons with weighted inputs, and a linear activation function. In other words, a standard FIR filter is a neural network. I mean, it's called &...
Marcus Müller's user avatar
3 votes
Accepted

2d convolution: What is the difference between convolution using blocked Toeplitz matrix and convolution layers?

The former is calculating a full sized convolution resulting in an N + M - 1 size, while the later is probably calculating a "valid" sized convolution. This case only returns the values from ...
Gillespie's user avatar
  • 1,767
2 votes
Accepted

Answering Machine vs. Human -- Neural Network Features Selection

I don't think that you would do this by feature extraction. Instead, I would train a many-to-one LSTM, which classifies the input sequence as -1 or 1. For you, it is easy to collect a huge amount of ...
Tolga Birdal's user avatar
  • 5,465
2 votes

Can deep neural networks achieve real-time video analysis?

Here is a newly published paper and video example: https://www.youtube.com/watch?v=w2iV8gt5cd4 http://arxiv.org/abs/1411.4389
Kon Rad's user avatar
  • 21
2 votes

Neural Network: Spectrogram Dimension

Depends on on what kind of feature(s) you want your CNN trained. With 2D convolution, convolution in the F dimension of a T-F spectrogram will produce something like a real cepstrum, which has proved ...
hotpaw2's user avatar
  • 35.3k
2 votes

What would the target matrix to train Neural Network?

Example with the following numbers (I use random numbers): There are 6 possible classes (face expressions in your case). You have 2000 examples (lets say 2000 photos of faces from which you know the ...
pablo_worker's user avatar
2 votes
Accepted

Does an equivalent transformation of a signal to a spectrogram image exist in which the phase information is part of the resulting image?

A homemade solution comes to my mind, but I don't know if it will work for you. I'll write it down anyway, since it may be helpful. In MATLAB you can do: ...
Tendero's user avatar
  • 5,020
2 votes

Does an equivalent transformation of a signal to a spectrogram image exist in which the phase information is part of the resulting image?

If you create a volumetric (unflattened 3D) image, you can use 2 layers in the 3rd dimension to represent magnitude and phase, or real and imaginary components of a complex spectrogram output. In a ...
hotpaw2's user avatar
  • 35.3k
2 votes

Deep learning model for phone recognition - issues with dimension the model

Since next layer is fully connected it does not really matter what shape your pooling output would be. You have 14x100, you can rearrange them as 1x1400 as input for next layer, 1000 elements as ...
Nikolay Shmyrev's user avatar
2 votes

Neural Networks and Complex Valued Inputs

Your question makes sense. The difference between a complex network and a regular network with twice the amount of channels, on a mechanical level, is the multiplication operation which ties pairs of ...
Roee Shenberg's user avatar
2 votes
Accepted

"Bi Directional" Kalman Filter - Kalman Filter for Smoothing

Anuar Y, Welcome to the DSP community. What you're talking about is called smoothing. Let me explain, assume we have samples $ {\left\{ x \left[ n \right] \right\}}_{n = 0}^{N - 1} $ and we want to ...
Royi's user avatar
  • 19.6k
2 votes

In Convolutional Neural Nets, what do convolutions look like?

The kernels used by a ConvNet are nothing but neural weights. You can think of them as a multilayer perceptron with some connections cut off and some weights restricted to be equal (weight sharing). ...
Tendero's user avatar
  • 5,020
2 votes

How to train and test deep neural network using MFCC features?

You may want to try training your model on 2D arrays containing entire datasets (all 75 MFCC frames of each audio file). You can try recurrent or convolutional early layers in your DNN to handle the ...
hotpaw2's user avatar
  • 35.3k
2 votes

Using STFT as an input to a Neural Net

Given a $M \times N$ STFT (spectrogram), use this as the input to a convolutional neural network. Do not flatten the spectrogram. Since your spectrogram will be complex, then you can use the magnitude ...
mhdadk's user avatar
  • 350
2 votes

Is a neural network an adaptive filter?

or is it called a neural network because it is "fancy"? Machine neural networks are called such because they deliberately emulate the functioning of biological neural networks, in an ...
TimWescott's user avatar
  • 12.7k
2 votes

Is a neural network an adaptive filter?

An adaptive filter is a special case of a neural network (NN). They have in common that they multiply an input x[n] with weights w[n], the result y[n]=x[n]w[n] is compared to the target t[n] (e.g. the ...
Aaron Verweg's user avatar
2 votes

Optimize window length (STFT) via gradient descent (in neural networks)

The easiest way is to take STFT using operators with autodiff support, e.g. via PyTorch. Then simply set the window as an updatable parameter, initializing as Hanning etc: ...
OverLordGoldDragon's user avatar
2 votes

Log of Filterbank Energies

Why is this done? Mostly because human auditory perception works this way. The relationship between energy and perceived loudness is logarithmic. That is not only true for hearing but for most other ...
Hilmar's user avatar
  • 44.6k
1 vote
Accepted

Generalized translation on graph

To understand either of these, you first have to understand the basic premise behind Graph Signal Processing (GSP) which is to map a signal to a graph and then work with it on the "Graph space". This ...
A_A's user avatar
  • 10.7k

Only top scored, non community-wiki answers of a minimum length are eligible