Hilbert space-filling curve is used for the generation of images of sEMG signals applied to CNNs.

What is it about?

We utilise the Hilbert space-filling curve for the generation of image representations of sEMG signals, which allows the application of typical image processing pipelines such as CNNs on sequence data. The proposed method is evaluated on different state-of-the-art network architectures and yields a significant classification improvement over the approach without the Hilbert curve. Additionally, we develop a new network architecture (MSHilbNet) that takes advantage of multiple scales of an initial Hilbert curve representation and achieves equal performance with fewer convolutional layers.

Featured Image

Why is it important?

This paper investigates the generation of image representations of sEMG using the Hilbert fractal curve. The proposed methodology offers an alternative for the classification of sEMG patterns using image processing methods, while using the Hilbert curve offers the advantage of locality preservation. Two methods (HilbTime and HilbElect) were evaluated and showed superior performance across various networks compared to the window segmentation method (baseline). However, the benefit was smaller for models with many parameters. Then, we presented a model (MSHilbNet) with few trainable parameters that utilizes multiple scales of the initial Hilbert curve representation. The evaluation of this multi-scale topology suggested that in every case it performed better than regular topologies based on VGGNet, DenseNet and Squee- zeNet. Finally, an analysis provided insights into the performance of the MSHilbNet architecture.