As attitude and motion sensing components, inertial sensors are widely used
in various portable devices. But the severe errors of inertial sensors restrain
their function, especially the trajectory recovery and semantic recognition. As
a mainstream signal processing method, wavelet is hailed as the mathematical
microscope of signal due to the plentiful and diverse wavelet basis functions.
However, complicated noise types and application scenarios of inertial sensors
make selecting wavelet basis perplexing. To this end, we propose a wavelet
dynamic selection network (WDSNet), which intelligently selects the appropriate
wavelet basis for variable inertial signals. In addition, existing deep
learning architectures excel at extracting features from input data but neglect
to learn the characteristics of target categories, which is essential to
enhance the category awareness capability, thereby improving the selection of
wavelet basis. Therefore, we propose a category representation mechanism (CRM),
which enables the network to extract and represent category features without
increasing trainable parameters. Furthermore, CRM transforms the common fully
connected network into category representations, which provide closer
supervision to the feature extractor than the far and trivial one-hot
classification labels. We call this process of imposing interpretability on a
network and using it to supervise the feature extractor the feature supervision
mechanism, and its effectiveness is demonstrated experimentally and
theoretically in this paper. The enhanced inertial signal can perform
impracticable tasks with regard to the original signal, such as trajectory
reconstruction. Both quantitative and visual results show that WDSNet
outperforms the existing methods. Remarkably, WDSNet, as a weakly-supervised
method, achieves the state-of-the-art performance of all the compared
fully-supervised methods.
Inertial sensors are widely used in various portable devices as attitude and motion sensing components. However, these sensors suffer from severe errors that restrict their functionality, particularly in areas such as trajectory recovery and semantic recognition. One popular signal processing method for dealing with these errors is wavelet analysis, which is known for its ability to dissect signals using various wavelet basis functions.
While wavelet analysis is a powerful tool, selecting the appropriate wavelet basis for variable inertial signals can be challenging due to the complex noise types and application scenarios involved. To address this issue, the authors of this paper propose a wavelet dynamic selection network (WDSNet) that intelligently selects the suitable wavelet basis for different inertial signals.
The WDSNet is designed to not only extract features from input data but also learn the characteristics of target categories. This is achieved through the use of a category representation mechanism (CRM), which allows the network to extract and represent category features without adding extra trainable parameters. By transforming the common fully connected network into category representations, the CRM provides closer supervision to the feature extractor, thereby enhancing the network’s category awareness capability.
The authors refer to this process of imposing interpretability on a network and using it to supervise the feature extractor as the feature supervision mechanism. They demonstrate the effectiveness of this mechanism experimentally and theoretically in their study.
The results of their experiments show that the enhanced inertial signal generated by WDSNet outperforms existing methods. It is able to perform impracticable tasks, such as trajectory reconstruction, which were not possible with the original signal. WDSNet also achieves state-of-the-art performance compared to fully-supervised methods, despite being a weakly-supervised method.
This research highlights the multi-disciplinary nature of the concepts explored, combining expertise in signal processing, machine learning, and sensor technology. The development of WDSNet and the feature supervision mechanism has the potential to improve the functionality of inertial sensors in portable devices, enabling better trajectory recovery and semantic recognition capabilities.