Search for: All records

Smart clothing has exhibited impressive body pose/movement tracking capabilities while preserving the soft, comfortable, and familiar nature of clothing. For practical everyday use, smart clothing should (1) be available in a range of sizes to accommodate different fit preferences, and (2) be washable to allow repeated use. In SeamFit, we demonstrate washable T-shirts, embedded with capacitive seam electrodes, available in three different sizes, for exercise logging. Our T-shirt design, customized signal processing & machine learning pipeline allow the SeamFit system to generalize across users, fits, and wash cycles. Prior wearable exercise logging solutions, which often attach a miniaturized sensor to a body location, struggle to track exercises that mainly involve other body parts. SeamFit T-shirt naturally covers a large area of the body and still tracks exercises that mainly involve uncovered joints (e.g., elbows and the lower body). In a user study with 15 participants performing 14 exercises, SeamFit detects exercises with an accuracy of 89%, classifies exercises with an accuracy of 93.4%, and counts exercises with an error of 0.9 counts, on average. SeamFit is a step towards practical smart clothing for everyday uses. 

Collective motion, which is ubiquitous in nature, has traditionally been explained by “self-propelled particle” models from theoretical physics. Here we show, through field, lab, and virtual reality experimentation, that classical models of collective behavior cannot account for how collective motion emerges in marching desert locusts, whose swarms affect the livelihood of millions. In contrast to assumptions made by these models, locusts do not explicitly align with neighbors. While individuals respond to moving-dot stimuli through the optomotor response, this innate behavior does not mediate social response to neighbors. Instead, locust marching behavior, across scales, can be explained by a minimal cognitive framework, which incorporates individuals’ neural representation of bearings to neighbors and internal consensus dynamics for making directional choices. Our findings challenge long-held beliefs about how order can emerge from disorder in animal collectives. 

Cross-lingual speech emotion recognition (SER) is important for a wide range of everyday applications. While recent SER research relies heavily on large pretrained models for emotion training, existing studies often concentrate solely on the final transformer layer of these models. However, given the task-specific nature and hierarchical architecture of these models, each transformer layer encapsulates different levels of information. Leveraging this hierarchical structure, our study focuses on the information embedded across different layers. Through an examination of layer feature similarity across different languages, we propose a novel strategy called a layer-anchoring mechanism to facilitate emotion transfer in cross-lingual SER tasks. Our approach is evaluated using two distinct language affective corpora (MSP-Podcast and BIIC-Podcast), achieving a best UAR performance of 60.21% on the BIIC-podcast corpus. The analysis uncovers interesting insights into the behavior of popular pretrained models. 

We present Ring-a-Pose, a single untethered ring that tracks continuous 3D hand poses. Located in the center of the hand, the ring emits an inaudible acoustic signal that each hand pose reflects differently. Ring-a-Pose imposes minimal obtrusions on the hand, unlike multi-ring or glove systems. It is not affected by the choice of clothing that may cover wrist-worn systems. In a series of three user studies with a total of 36 participants, we evaluate Ring-a-Pose's performance on pose tracking and micro-finger gesture recognition. Without collecting any training data from a user, Ring-a-Pose tracks continuous hand poses with a joint error of 14.1mm. The joint error decreases to 10.3mm for fine-tuned user-dependent models. Ring-a-Pose recognizes 7-class micro-gestures with a 90.60% and 99.27% accuracy for user-independent and user-dependent models, respectively. Furthermore, the ring exhibits promising performance when worn on any finger. Ring-a-Pose enables the future of smart rings to track and recognize hand poses using relatively low-power acoustic sensing. 

The prevalence of cross-lingual speech emotion recognition (SER) modeling has significantly increased due to its wide range of applications. Previous studies have primarily focused on technical strategies to adapt features, domains, and labels across languages, often overlooking the underlying universalities between the languages. In this study, we address the language adaptation challenge in cross-lingual scenarios by incorporating vowel-phonetic constraints. Our approach is structured in two main parts. Firstly, we investigate the vowel-phonetic commonalities associated with specific emotions across languages, particularly focusing on common vowels that prove to be valuable for SER modeling. Secondly, we utilize these identified common vowels as anchors to facilitate cross-lingual SER. To demonstrate the effectiveness of our approach, we conduct case studies using American English, Taiwanese Mandarin, and Russian using three naturalistic emotional speech corpora: the MSP-Podcast, BIIC-Podcast, and Dusha corpora. The proposed unsupervised cross-lingual SER model, leveraging this phonetic information, surpasses the performance of the baselines. This research provides insights into the importance of considering phonetic similarities across languages for effective language adaptation in cross-lingual SER scenarios.