Papers for

healthcare monitoring teams

Papers whose findings have a practical use for this group, as judged from the abstract. Open a paper to read what it means in practice.

Tempo extends language models to understand time series data

TEMPO: Time Series Understanding via Discrete Tokenization

Abstract: Time-series interpretation often requires both numerical observations and textual context. Integrating these inputs into a pretrained language model raises a representation question: how should a signal enter the model's token sequence? We present TEMPO, which extends an LLM's vocabulary with discrete codes from a context-aware Transformer tokenizer using finite scalar quantization. A frozen tokenizer maps four-step patches to a 625-code vocabulary; signal tokens, scale statistics, and text are then processed by the LLM's existing self-attention layers. Adaptation updates the new token embeddings and DoRA adapters. With a Qwen3-4B backbone and task-specific fine-tuning, TEMPO achieves macro-F1 scores of 89.0% on activity recognition, 85.0% on sleep staging, and 86.0% on bearing-fault diagnosis. Time-series question answering and qualitative bearing-diagnosis examples further illustrate how this interface combines signal representations with textual context. Together, these results demonstrate the utility of discrete vocabulary extension for time-series understanding.

Tue 3 NovComputation and Language
The gist
Interpreting time series data, like activity or sleep patterns, is tricky because it mixes numbers and words. The authors developed TEMPO, which turns chunks of time series data into special codes that a language model can understand alongside text. This approach helps the model make sense of both signals and written context together. They tested TEMPO on tasks like recognizing activities and diagnosing faults, showing it works well.
Open → 2611.04482v1

Molecular sensors detect artery plaque using particle timing patterns

Molecular Communication for Plaque Sensing in a Vascular Bifurcation with Receiver Constraints

Abstract: Future molecular communication (MC) networks are envisioned to use the human cardiovascular system as a channel. Disease-related changes affect particle transport, allowing an MC receiver to also act as a sensor. In this work, we investigate plaque formation in a carotid bifurcation through a systematic fluid-dynamics simulation study with non-Newtonian blood rheology. We analyze how plaque geometry changes branch-specific particle reception and timing, and use these observations to estimate plaque-induced channel obstruction. At an illustrative reference of only 128 emitted particles, a 250 ms observation window and 50 ms temporal resolution, a Ridge detector estimates maximum local channel-width reduction with a mean absolute error of 8.30 percentage points. We evaluate robustness across system constraints like particle budgets, observation windows, branch access, detection efficiency, count resolution, and synchronization. Coarse counting has little effect when timing features remain available, whereas limited temporal resolution and clock uncertainty substantially affect short-window estimates. Extending the observation window to 1 s reduces these timing-related penalties. The results establish a baseline for plaque sensing from constrained MC observations and identify receiver capabilities relevant to the Internet of Bio-Nano Things.

Mon 2 NovEmerging Technologies
The gist
This work studies how tiny particles can be used to sense blockages caused by plaque in blood vessels, especially at points where vessels split. The researchers used computer simulations of blood flow to see how plaque changes how particles travel and arrive at sensors. They found ways to estimate the amount of blockage based on when and how many particles a sensor detects, even with limited data and timing accuracy. This approach could help build tiny communication networks inside the body that also monitor health conditions.
Open → 2611.02773v1

MURFormer improves heart rate signal quality by reducing noise

MURFormer: Attention-Based Lightweight Multi-Scale Uncertainty-Guided Refinement Framework for PPG Denoising

Abstract: Photoplethysmography (PPG) is a widely used physiological signal for its convenience but is highly susceptible to motion artifacts that degrade waveform quality and limit downstream analysis. This paper proposes MURFormer, a lightweight multi-scale uncertainty-guided refinement framework for PPG denoising through noisy-to-clean PPG mapping. MURFormer employs a newly proposed Multi-Scale Attention-Conv Fusion (MACF) block, which is designed to efficiently capture multi-scale temporal dynamics of PPG, as the fundamental building block and performs uncertainty-guided two-pass refinement prioritizing unreliable regions. The method was evaluated on the publicly available CIME dataset and on a Daily Activities dataset collected in this work, comprising paired noisy-clean PPG recordings from 24 subjects. Experimental results show that MURFormer improves both signal quality and physiological fidelity over the raw noisy input, yielding 129\% and 105\% relative signal-to-noise ratio (SNR) improvements, and 66\% and 81\% relative reductions in heart-rate mean absolute error (MAE), on the Daily Activities and CIME datasets, respectively. MURFormer also outperforms an existing denoising baseline on both datasets with a relatively small parameter count while preserving beat-level timing and waveform morphology. In addition, the model maintains strong denoising performance across varying input window lengths, suggesting that MURFormer is an effective and robust framework for practical PPG denoising. The implementation of MURFormer is publicly available at https://github.com/k-hram/MURFormer.git

Mon 2 NovArtificial IntelligenceMachine Learning
The gist
Photoplethysmography (PPG) measures heart signals easily but can get messy when people move, making the data hard to use. The authors propose MURFormer, a lightweight computer method that cleans noisy PPG signals by focusing on uncertain parts and refining them twice. They tested MURFormer on two datasets with real noisy and clean heart signal pairs and found it greatly improved signal clarity and accuracy in measuring heart rates. It also works well on different time lengths of data and keeps important heartbeat details intact.
Open → 2611.02715v1

BrainCodes offers compact EEG model for diverse brain monitoring tasks

BrainCodes: A Lightweight EEG Foundation Model over Discrete Neural Codes

Abstract: Recent electroencephalography (EEG) foundation models increasingly emphasize model scale, potentially limiting their deployment on resource-constrained devices and in everyday environments. We argue that transferable EEG features can instead be learned with a compact architecture and introduce BrainCodes, a lightweight foundation model pretrained through self-supervised learning on more than 800 hours of recordings from over 1,000 participants. Our architecture combines a quantization-based neural tokenizer with a spatiotemporal encoder that supports arbitrary electrode configurations. The pretraining objective explicitly encourages representations that capture spatial connectivity through channel positional encoding, and sensitivity to narrowband neural oscillations through an auxiliary training objective. Our model matches or surpasses the state of the art across diverse neurophysiological monitoring tasks, including sleep staging, attention assessment, and cognitive load estimation. BrainCodes transfers effectively to low-resource wearable devices while using up to 97% fewer trainable parameters than comparable architectures. Beyond predictive accuracy, our evaluation framework characterizes the geometry and invariances of learned EEG representations across measurable spatial, temporal, and spectral signal properties.

Wed 21 OctMachine LearningNeural and Evolutionary Computing
The gist
Analyzing brain signals with EEG often needs big and complex models, which can be hard to use on small devices. The authors created BrainCodes, a small but powerful model trained on many hours of brain recordings from lots of people. This model learns useful brain features that work well for tasks like tracking sleep stages and attention, and can run on devices with limited computing power. It combines special ways to handle signals from different brain locations and frequencies.
Open → 2610.29320v1

LineupRL improves time series captioning using verifiable reinforcement learning

LineupRL: Verifiable Reinforcement Learning for Time Series Captioning via Caption-to-Series Identification

Abstract: Time series captioning is a fundamental step in time series understanding and can also serve as the bridge between signal and natural language. Supervised fine-tuning (SFT) relies on a larger model's captions and cannot exceed their quality. Reinforcement learning (RL) can, but its rewards were designed for other modalities and other tasks, and they transfer poorly to open-ended generation in the time series domain. We address this by proposing LineupRL, a reinforcement learning with verifiable rewards (RLVR) pipeline whose reward is caption-to-series identification. The reward model is a frozen large language model (LLM) verifier that reads the generated caption and the candidate time series as raw values, never the chart, and must pick the described time series from multiple distractors. Matching is a far lighter demand on the verifier than writing questions or judging a caption, so an off-the-shelf LLM can supply the reward. Across two captioning benchmarks, and on forecasting and reconstruction where the predictor sees only the caption, LineupRL outperforms SFT and RL baselines on every metric. The 3B vision language model (VLM) trained by LineupRL also outperforms, at 1/24 of the parameters, the 72B VLM whose captions the SFT baseline is distilled from. Our case study shows that LineupRL resists reward hacking, and that the captioner it trains both traces the trend and names the values at key points.

Thu 1 OctArtificial Intelligence
The gist
Understanding time series data often involves creating captions that describe patterns and important points. The authors propose a new approach called LineupRL that uses reinforcement learning with a special reward system. This reward checks if a generated caption can help a language model correctly identify the matching time series among distractors. Their method performs better than previous approaches and creates captions that more accurately describe the data.
Open → 2610.01800v1

Human activity recognition framework improves wearable sensor accuracy

An Effective, Reliable, and Robust Framework for Human Activity Recognition Using Wearable Sensors

Abstract: Human Activity Recognition (HAR) through wearable sensors greatly improves the quality of human life through its multiple applications. For HAR, multi-sensor channel information is vital for optimal performance. Current work states that applying an attention neural network to prioritize discriminatory sensor channels helps the model classify activity more precisely. However, obtaining discriminatory information from multisensory channels is not always trivial, such as when collecting data from older hospitalized patients. In this context, existing HAR methods struggle to classify activities, particularly activities with similar natures. Moreover, HAR models predominantly suffer from overfitting due to the small size of available datasets, which leads to poor performance. Data augmentation (DA) is a viable solution to this problem. However, available DA methods have various drawbacks, including the possibility of being domain-dependent, resulting in distorted models for test sequences. To address these HAR problems, we propose a novel framework, ALAE-TAE-CutMix+, which focuses on two aspects. First, it enhances the latent information across each sensor channel and learns to exploit the relation among multiple latent features and the ongoing activity. Consequently, the discriminatory feature representations of each activity is enriched. Second, a new augmentation strategy is introduced to address the shortcomings of existing multi-sensor channel data augmentation. We then extend the framework to create a further enhanced version, namely ALAE-CIE-TAE-CutMix+, which learns to capture the interactions between the features of each pair of sensor channels. We find that although the first framework performs slightly better than the latter, the latter is nonetheless more reliable and robust. Both frameworks significantly outperform SOTA approaches on the four HAR datasets from diverse domains.

Tue 29 SeptMachine Learning
The gist
Recognizing human activities using wearable sensors can help improve health monitoring and daily life assistance. The authors found that existing methods struggle when data is limited or when activities are very similar, especially for patients like older adults. They created a new method that better uses information from multiple sensors and improves data variety without distorting it. Their approach works more reliably and accurately across several different activity datasets than previous methods.
Open → 2609.36848v1

Time series generation improved with hyperspherical latents and masked modeling

HALO: Enhancing Time Series Generation via Hyperspherical Latents and Masked AutoregRessive Modeling

Abstract: Most existing time series generators rely on a two-stage modeling paradigm: the first stage learns discrete latent representations of time series; the second stage performs autoregressive modeling on these discrete latents through next token prediction. However, this paradigm suffers from two stage-specific limitations: the first stage can lead to information loss when discretizing continuous time series, while the second stage is prone to error accumulation during autoregressive generation. To address these limitations, our core idea is to perform generative modeling in a continuous latent space with a more efficient autoregressive framework. We propose HALO, which enhances time series generation via Hyperspherical Latents and Masked Autoregressive modeling to achieve this goal by tackling two key bottlenecks: (1) variance and scale heterogeneity of continuous latent representations; (2) the difficulty of balancing generation efficiency with temporal correlation modeling. HALO first introduces a hyperspherical VAE that constrains continuous latents to a fixed-radius hyperspherical shell, effectively stabilizing the numerical fluctuations of continuous latent representations. Secondly, we develop a masked autoregressive model that balances parallel decoding and temporal correlation learning, substantially reducing the number of inference steps required for generation and improving generation stability. Our extensive experiments demonstrate that HALO achieves state-of-the-art generation performance while offering significantly improved inference efficiency over existing advanced baselines.

Mon 28 SeptMachine Learning
The gist
Generating realistic time series data is hard because current methods lose important details or make compounding errors when producing new data. The authors present HALO, a method that represents time series in a stable, continuous space shaped like a sphere, reducing information loss. HALO also uses a smarter way to generate sequences that better balances speed and accuracy. This results in more reliable and efficient time series generation compared to previous approaches.
Open → 2609.34511v1