Multichannel Displacement measurement via self mixing interferometry and neural network : training and test datasets

Self mixing interferometry is a simple and robust sensing method which can be used (among other things) to measure the displacement of a target along the light propagation axis. While conceptually simple, the actual use of this method is less straightforward than originally envisioned because reconstructing the target displacement from the interferometric signal is often tricky. A small neural network can do this task very well after proper training, as described in [10.1364/OE.419844], with dataset [10.5281/zenodo.7303745]. 

Here, the dataset is composed by a training set and a test set, in a specific configuration in which 3 self-mixing sensors measure simultaneously the same target displacement. Both datasets contain the displacement itself and the 3 interferometric signals (1 per sensing channel)

The training set relies on two python/numpy data files corresponding to harmonic displacements for different frequencies ranging from 53 and 93 Hz and amplitudes from 3.5 to 7.5 µm : 

Training_set_2_lostchannel_displacement.npy : 93744-elements long numpy array containing the target's displacement in units of µm/ms with a 1.024 ms time step. 

Training_set_2_lostchannel_signal.npy : numpy array of shape (3, 93744, 256, 1) containing the interferometric signals. The first dimension refers to the channel (1, 2 or 3), the second dimension is the number of segments of 256 points. Each segment of 256 points correspond to a 1.024 ms window of signal, matching one element of the displacement. For instance, the displacement value in `displacement[416]` corresponds to the interferometric signal segment `signal[0,416,:,0]` for channel 1, `signal[1,416,:,0]` for channel 2 and `signal[2,416,:,0]` for channel 3. 

The test set follows the same architecture and format as the training set, but contains only random displacements generated by a delta-correlated signal, which we Fourier filter with a fifth order Butterworth filter between 10 and 100 Hz :

"Displacement_test.npy" :  with shape (246078, 1)

"Signal_test.npy" : with shape (3, 246078, 256, 1)

Only the displacement type has changed from harmonic to random, from the training to the test datasets.

These datasets have been used to train and test a 3 channel neural network (after data augmentation) in order to emphasize the high availability potential of multichannel schemes, against backscattered power fluctuations.