Leveraging multimodal information from biosignals is vital for building a comprehensive representation of people's physical and mental states. However, multimodal biosignals often exhibit substantial distributional shifts between pretraining and inference datasets, stemming from changes in task specification or variations in modality compositions. To achieve effective pretraining in the presence of potential distributional shifts, we propose a frequency-aware masked autoencoder ($\texttt{bio}$FAME) that learns to parameterize the representation of biosignals in the frequency space. $\texttt{bio}$FAME incorporates a frequency-aware transformer, which leverages a fixed-size Fourier-based operator for global token mixing, independent of the length and sampling rate of inputs. To maintain the frequency components within each input channel, we further employ a frequency-maintain pretraining strategy that performs masked autoencoding in the latent space. The resulting architecture effectively utilizes multimodal information during pretraining, and can be seamlessly adapted to diverse tasks and modalities at test time, regardless of input size and order. We evaluated our approach on a diverse set of transfer experiments on unimodal time series, achieving an average of $\uparrow$5.5% improvement in classification accuracy over the previous state-of-the-art. Furthermore, we demonstrated that our architecture is robust in modality mismatch scenarios, including unpredicted modality dropout or substitution, proving its practical utility in real-world applications. Code will be available soon.
Accurate recognition of hand gestures is crucial to the functionality of smart prosthetics and other modern human-computer interfaces. Many machine learning-based classifiers use electromyography (EMG) signals as input features, but they often misclassify gestures performed in different situational contexts (changing arm position, reapplication of electrodes, etc.) or with different effort levels due to changing signal properties. Here, we describe a learning and classification algorithm based on hyperdimensional (HD) computing that, unlike traditional machine learning algorithms, enables computationally efficient updates to incrementally incorporate new data and adapt to changing contexts. EMG signal encoding for both training and classification is performed using the same set of simple operations on 10,000-element random hypervectors enabling updates on the fly. Through human experiments using a custom EMG acquisition system, we demonstrate 88.87% classification accuracy on 13 individual finger flexion and extension gestures. Using simple model updates, we preserve this accuracy with less than 5.48% degradation when expanding to 21 commonly used gestures or when subject to changing situational contexts. We also show that the same methods for updating models can be used to account for variations resulting from the effort level with which a gesture is performed.
EMG-based gesture recognition shows promise for human-machine interaction. Systems are often afflicted by signal and electrode variability which degrades performance over time. We present an end-to-end system combating this variability using a large-area, high-density sensor array and a robust classification algorithm. EMG electrodes are fabricated on a flexible substrate and interfaced to a custom wireless device for 64-channel signal acquisition and streaming. We use brain-inspired high-dimensional (HD) computing for processing EMG features in one-shot learning. The HD algorithm is tolerant to noise and electrode misplacement and can quickly learn from few gestures without gradient descent or back-propagation. We achieve an average classification accuracy of 96.64% for five gestures, with only 7% degradation when training and testing across different days. Our system maintains this accuracy when trained with only three trials of gestures; it also demonstrates comparable accuracy with the state-of-the-art when trained with one trial.