Human Activity Recognition (HAR) has been studied extensively, yet current approaches are not capable of generalizing across different domains (i.e., subjects, devices, or datasets) with acceptable performance. This lack of generalization hinders the applicability of these models in real-world environments. As deep neural networks are becoming increasingly popular in recent work, there is a need for an explicit comparison between handcrafted and deep representations in Out-of-Distribution (OOD) settings. This paper compares both approaches in multiple domains using homogenized public datasets. First, we compare several metrics to validate three different OOD settings. In our main experiments, we then verify that even though deep learning initially outperforms models with handcrafted features, the situation is reversed as the distance from the training distribution increases. These findings support the hypothesis that handcrafted features may generalize better across specific domains.

The study of Domain Generalization (DG) has gained considerable momentum in the Machine Learning (ML) field. Human Activity Recognition (HAR) inherently encompasses diverse domains (e.g., users, devices, or datasets), rendering it an ideal testbed for exploring Domain Generalization. Building upon recent work, this paper investigates the application of regularization methods to bridge the generalization gap between traditional models based on handcrafted features and deep neural networks. We apply various regularizers, including sparse training, Mixup, Distributionally Robust Optimization (DRO), and Sharpness-Aware Minimization (SAM), to deep learning models and assess their performance in Out-of-Distribution (OOD) settings across multiple domains using homogenized public datasets. Our results show that Mixup and SAM are the best-performing regularizers. However, they are unable to match the performance of models based on handcrafted features. This suggests that while regularization techniques can improve OOD robustness to some extent, handcrafted features remain superior for domain generalization in HAR tasks.

Objectives: Cough dysfunction is a feature of patients with amyotrophic lateral sclerosis (ALS). The cough sounds carry information about the respiratory system and bulbar involvement. Our goal was to explore the association between cough sound characteristics and the respiratory and bulbar functions in ALS. Methods: This was a single-center, cross-sectional, and case-control study. On-demand coughs from ALS patients and healthy controls were collected with a smartphone. A total of 31 sound features were extracted for each cough recording using time-frequency signal processing analysis. Logistic regression was applied to test the differences between patients and controls, and in patients with bulbar and respiratory impairment. Support vector machines (SVM) were employed to estimate the accuracy of classifying between patients and controls and between patients with bulbar and respiratory impairment. Multiple linear regressions were applied to examine correlations between cough sound features and clinical variables. Results: Sixty ALS patients (28 with bulbar dysfunction, and 25 with respiratory dysfunction) and forty age- and gender-matched controls were recruited. Our results revealed clear differences between patients and controls, particularly within the frequency-related group of features (AUC 0.85, CI 0.79–0.91). Similar results were observed when comparing patients with and without bulbar dysfunction. Sound features related to intensity displayed the strongest correlation with disease severity, and were the most significant in distinguishing patients with and without respiratory dysfunction. Discussion: We found a good relationship between specific cough sound features and clinical variables related to ALS functional disability. The findings relate well with some expected impact from ALS on both respiratory and bulbar contributions to the physiology of cough. Finally, our approach could be relevant for clinical practice, and it also facilitates home-based data collection.

Objectives Cough dysfunction is a feature of patients with amyotrophic lateral sclerosis (ALS). The cough sounds carry information about the respiratory system and bulbar involvement. Our goal was to explore the association between cough sound characteristics and the respiratory and bulbar functions in ALS. Methods This was a single-center, cross-sectional, and case-control study. On-demand coughs from ALS patients and healthy controls were collected with a smartphone. A total of 31 sound features were extracted for each cough recording using time-frequency signal processing analysis. Logistic regression was applied to test the differences between patients and controls, and in patients with bulbar and respiratory impairment. Support vector machines (SVM) were employed to estimate the accuracy of classifying between patients and controls and between patients with bulbar and respiratory impairment. Multiple linear regressions were applied to examine correlations between cough sound features and clinical variables. Results Sixty ALS patients (28 with bulbar dysfunction, and 25 with respiratory dysfunction) and forty age- and gender-matched controls were recruited. Our results revealed clear differences between patients and controls, particularly within the frequency-related group of features (AUC 0.85, CI 0.79- 0.91). Similar results were observed when comparing patients with and without bulbar dysfunction; and with and without respiratory dysfunction. Sound features related to intensity displayed the strongest correlation with disease severity. Discussion We found a good relationship between specific cough sound features and clinical variables related to ALS functional disability. The findings relate well with some expected impact from ALS on both respiratory and bulbar contributions to the physiology of cough. Finally, our approach could be relevant for clinical practice, and it also facilitates home-based data collection.