What type of neural network is best for decoding time sequences?

Question

Neural networks

And are ANNs generally suitable for such tasks?
Let's assume that the values ​​of signal durations are applied to the network inputs (the signal is fixed and consists of 6 durations and intervals), the durations of these intervals and pulses for each signal are different, but in total they always give one value. If you train the network on the correct and incorrect values ​​of these signals, will it be able to determine signals that are close to the truth and reject the "noise"? If so, what types of neural networks are suitable for this task? I tried to assemble a one / two-layer perceptron, but either the arms are crooked, or this structure does not fit, in general, it turned out to be nonsense.
Examples of signals (vectors supplied) are:
0.15 + 0.2 + 0.3 + 0.1 + 0.1 + 0.15 = 1 sec - "good" signal
0.2 + 0.1 + 0.25 + 0.2 + 0.1 + 0.15 = 1 sec - "bad" signal
0.3 + 0.2 + 0 + 0.1 + 0.2 + 0, 2 = 1sec - "good" signal
, etc.
I trained the network by giving "good" and composing "bad" signals, replacing the necessary digits with zeros/other digits in some places.
I understand that this can be implemented with conventional algorithms (which I did), but I would like to try such an unusual method.
Thanks to all who responded! :)

Answer 1

NS can do anything. This is a universal approximating function.
And on the topic, DSP will help you as a general discipline, and the signal correlation function as a private hint.
Don't bother with neural networks. It is a scientifically proven fact that they can do everything, but there will always be a more specialized method that will be better than NA.