Yang Liu
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Research Subjects
Yang Liu has a total of 67 co-authors. Here are the top co-authors by number of shared publications:
- Cecilia Mascolo
Department of Computer Science and Technology, University of Cambridge, Cambridge, United Kingdom
14 shared publications - Zhenjiang Li
City University of Hong Kong
12 shared publications - Dong Ma
University of Cambridge, Cambridge, United Kingdom
10 shared publications - Jake Stuchbury-Wass
Department of Computer Science and Technology, University of Cambridge, Cambridge, United Kingdom
9 shared publications - Kayla-Jade Butkow
Department of Computer Science and Technology, University of Cambridge, Cambridge, United Kingdom
9 shared publications - Qiang Yang
Department of Computer Science and Technology, University of Cambridge, Cambridge, United Kingdom
9 shared publications
Latest Publications
- Open Access
- 30 September 2026
- Published by Association for Computing Machinery (ACM) in Proceedings of the ACM on Interactive, Mobile, Wearable and Ubiquitous Technologies
- Vol. 10 (3) , 1-25
ImpactEar: Cross Activity Ground Reaction Force Estimation using Earable IMUs
Ground Reaction Forces (GRFs) are the forces exerted between the foot and the ground during movement. They are key biomechanical indicators for gait monitoring, injury recovery tracking, and sporting performance. Traditional force plates provide gold-standard accuracy but are expensive and confined to laboratories, while leg- or foot-mounted devices require specialised hardware and sacrifice user comfort. Recent work utilising commodity devices such as smartwatches and earphones offer greater accessibility, yet wrist sensors suffer from arm-swing artifacts and existing earable studies remain activity-specific, limited to either walking or running (and never both) as GRF patterns differ drastically across locomotion types. In this paper, we propose ImpactEar, the first system to estimate complete GRF curves across multiple activities using only a pair of earable inertial measurement units (IMUs). Our core insight is to treat the head as a proxy for the body's center of mass and leverage bilateral earable IMUs to capture both translational and rotational dynamics. Combined with temporal context modeling, this enables a cross-activity mapping from head motion to GRF curves with a single model. ImpactEar employs a lightweight encoder-decoder network that reconstructs left-right GRF profiles across walking, running, and jumping in real time. The evaluation on 30 participants shows that ImpactEar achieves 8.6% normalized root mean square error (NRMSE) and 9.5 ms latency on a mobile phone streaming earable IMU data, outperforming state-of-the-art single-activity baselines and remaining robust under practical conditions, such as reduced sampling rates, single earbuds, different ground surfaces, music playback and head and facial motions. We further demonstrate the power of a cross activity model over a range of different downstream applications including gait asymmetry analysis, jump power assessment, and ballet motion evaluation. Finally, we release ImpactEarDS, the first public earable-GRF dataset with synchronised IMU and force-plate ground truth, paving the way for ubiquitous human kinetics and biomechanics research.
- 30 September 2026
- Published by Association for Computing Machinery (ACM) in Proceedings of the ACM on Interactive, Mobile, Wearable and Ubiquitous Technologies
- Vol. 10 (3) , 1-25
Open AccessImpactEar: Cross Activity Ground Reaction Force Estimation using Earable IMUs
Ground Reaction Forces (GRFs) are the forces exerted between the foot and the ground during movement. They are key biomechanical indicators for gait monitoring, injury recovery tracking, and sporting performance. Traditional force plates provide gold-standard accuracy but are expensive and confined to laboratories, while leg- or foot-mounted devices require specialised hardware and sacrifice user comfort. Recent work utilising commodity devices such as smartwatches and earphones offer greater accessibility, yet wrist sensors suffer from arm-swing artifacts and existing earable studies remain activity-specific, limited to either walking or running (and never both) as GRF patterns differ drastically across locomotion types. In this paper, we propose ImpactEar, the first system to estimate complete GRF curves across multiple activities using only a pair of earable inertial measurement units (IMUs). Our core insight is to treat the head as a proxy for the body's center of mass and leverage bilateral earable IMUs to capture both translational and rotational dynamics. Combined with temporal context modeling, this enables a cross-activity mapping from head motion to GRF curves with a single model. ImpactEar employs a lightweight encoder-decoder network that reconstructs left-right GRF profiles across walking, running, and jumping in real time. The evaluation on 30 participants shows that ImpactEar achieves 8.6% normalized root mean square error (NRMSE) and 9.5 ms latency on a mobile phone streaming earable IMU data, outperforming state-of-the-art single-activity baselines and remaining robust under practical conditions, such as reduced sampling rates, single earbuds, different ground surfaces, music playback and head and facial motions. We further demonstrate the power of a cross activity model over a range of different downstream applications including gait asymmetry analysis, jump power assessment, and ballet motion evaluation. Finally, we release ImpactEarDS, the first public earable-GRF dataset with synchronised IMU and force-plate ground truth, paving the way for ubiquitous human kinetics and biomechanics research.
- Open Access
- 17 July 2026
- Published by Springer Nature in Nature Communications
- Vol. 17 (1)
Measuring cardiac stroke volume through in-ear audio sensing
Stroke volume, the volume of blood ejected by the left ventricle during a contraction, is a key metric of cardiovascular health. Currently, stroke volume is measured in clinic with specialised equipment. While purpose-made wearables exist to measure stroke volume, no solution relies solely on commodity devices. We present a deep learning system for stroke volume estimation from in-ear audio of earbuds. We combine generative self-supervised/transfer learning, a transformer-based autoencoder, to predict average stroke volume in unseen subjects. With data from 23 healthy participants, we compare our estimations to clinically validated device estimations. We achieve a mean absolute error of 5.24 ml, a Pearson correlation of r=0.94 between average predicted stroke volume and average true stroke volume, and a Percentage Error in the limits of agreement of 11.05% (within clinical range for stroke volume measurement devices). These findings open the doors to longitudinal, scalable and affordable cardiovascular measurement out of clinic.
- 17 July 2026
- Published by Springer Nature in Nature Communications
- Vol. 17 (1)
Open AccessMeasuring cardiac stroke volume through in-ear audio sensing
Stroke volume, the volume of blood ejected by the left ventricle during a contraction, is a key metric of cardiovascular health. Currently, stroke volume is measured in clinic with specialised equipment. While purpose-made wearables exist to measure stroke volume, no solution relies solely on commodity devices. We present a deep learning system for stroke volume estimation from in-ear audio of earbuds. We combine generative self-supervised/transfer learning, a transformer-based autoencoder, to predict average stroke volume in unseen subjects. With data from 23 healthy participants, we compare our estimations to clinically validated device estimations. We achieve a mean absolute error of 5.24 ml, a Pearson correlation of r=0.94 between average predicted stroke volume and average true stroke volume, and a Percentage Error in the limits of agreement of 11.05% (within clinical range for stroke volume measurement devices). These findings open the doors to longitudinal, scalable and affordable cardiovascular measurement out of clinic.
- Open Access
- 20 June 2026
- Published by Association for Computing Machinery (ACM)
- p. 548-562
Reliable Metal Foreign Object Detection for Mobile Wireless Charging via Harmonic Fingerprinting
- 20 June 2026
- Published by Association for Computing Machinery (ACM)
- p. 548-562
Open AccessReliable Metal Foreign Object Detection for Mobile Wireless Charging via Harmonic Fingerprinting
- Open Access
- 15 June 2026
- Published by Association for Computing Machinery (ACM) in Proceedings of the ACM on Interactive, Mobile, Wearable and Ubiquitous Technologies
- Vol. 10 (2) , 1-41
A Survey of Earable Technology: Trends, Tools, and the Road Ahead
Earable devices, wearables positioned in or around the ear, are undergoing a rapid transformation from audio-centric accessories into multifunctional systems for interaction, contextual awareness, and health monitoring. This evolution is driven by commercial trends emphasizing sensor integration and by a surge of academic interest exploring novel sensing capabilities. Building on the foundation established by earlier surveys, this work presents a timely and comprehensive review of earable research published since 2022. We attempt to answer three core questions: (1) how has earable research evolved in recent years, (2) what enabling resources are now available, and (3) what opportunities remain for future exploration. For question (1), we analyze over one hundred recent studies to characterize this shifting research landscape, by identifying emerging sensing principles / applications, and efforts to improve accuracy and reliability over prior works. For question (2), we summarize new public datasets and hardware platforms that can be leveraged to facilitate future earable research. Finally, for question (3), we discuss open challenges and propose future directions for the next phase of earable research.
- 15 June 2026
- Published by Association for Computing Machinery (ACM) in Proceedings of the ACM on Interactive, Mobile, Wearable and Ubiquitous Technologies
- Vol. 10 (2) , 1-41
Open AccessA Survey of Earable Technology: Trends, Tools, and the Road Ahead
Earable devices, wearables positioned in or around the ear, are undergoing a rapid transformation from audio-centric accessories into multifunctional systems for interaction, contextual awareness, and health monitoring. This evolution is driven by commercial trends emphasizing sensor integration and by a surge of academic interest exploring novel sensing capabilities. Building on the foundation established by earlier surveys, this work presents a timely and comprehensive review of earable research published since 2022. We attempt to answer three core questions: (1) how has earable research evolved in recent years, (2) what enabling resources are now available, and (3) what opportunities remain for future exploration. For question (1), we analyze over one hundred recent studies to characterize this shifting research landscape, by identifying emerging sensing principles / applications, and efforts to improve accuracy and reliability over prior works. For question (2), we summarize new public datasets and hardware platforms that can be leveraged to facilitate future earable research. Finally, for question (3), we discuss open challenges and propose future directions for the next phase of earable research.
- Open Access
- 10 May 2026
- Published by Association for Computing Machinery (ACM)
- p. 615-621
Short Paper: EarSleeve: Transforming Everyday Earphones into a 12-Lead ECG Sensing Platform
- 10 May 2026
- Published by Association for Computing Machinery (ACM)
- p. 615-621
Open AccessShort Paper: EarSleeve: Transforming Everyday Earphones into a 12-Lead ECG Sensing Platform
- Open Access
- 10 May 2026
- Published by Association for Computing Machinery (ACM)
- p. 588-601
NutriEar: Robust Nutrition-Aware Food Classification from In-Ear Acoustic Signals
- 10 May 2026
- Published by Association for Computing Machinery (ACM)
- p. 588-601
Open AccessNutriEar: Robust Nutrition-Aware Food Classification from In-Ear Acoustic Signals
- Open Access
- 10 May 2026
- Published by Association for Computing Machinery (ACM)
- p. 661-674
NeuroPath: Practically Adopting Motor Imagery Decoding through EEG Signals
Motor Imagery (MI) is an emerging Brain–Computer Interface (BCI) paradigm in which a person imagines a body movement without any physical action. By decoding the scalp-recorded electroencephalography (EEG) signals, BCIs can establish direct communication pathways to control external devices, offering significant potential in prosthetics, rehabilitation, and human–computer interaction. However, existing solutions remain difficult to deploy in practice. (i) Most employ independent, opaque models for each MI task. This fragmented methodology lacks a unified architectural foundation. Consequently, these models are trained in isolation and fail to learn robust representations from diverse datasets, which often results in modest performance. (ii) They primarily adopt fixed sensor deployment, whereas real-world setups vary in electrode number and placement, causing models trained on one configuration to fail under another. (iii) Performance degrades sharply under low-SNR conditions typical of consumer-grade EEG. Together, these limitations hinder the practical adoption of MI-based BCIs. To address these challenges, we present NeuroPath, a well-designed neural architecture for robust MI decoding. NeuroPath takes high-level inspiration from the brain’s signal pathway from cortex to scalp, utilizing a deep neural architecture with specialized modules for signal filtering, spatial representation learning, and feature classification, enabling unified decoding rather than task-specific black-box learning. To handle variations in electrode configurations, we introduce a spatially aware graph adapter that accommodates different electrode numbers and placements. To enhance robustness under low-SNR conditions, NeuroPath incorporates multimodal auxiliary training to refine EEG representations and stabilize performance on noisy, real-world data. Evaluations on three self-collected consumer-grade datasets and three public medical-grade datasets demonstrate that NeuroPath achieves superior performance.
- 10 May 2026
- Published by Association for Computing Machinery (ACM)
- p. 661-674
Open AccessNeuroPath: Practically Adopting Motor Imagery Decoding through EEG Signals
Motor Imagery (MI) is an emerging Brain–Computer Interface (BCI) paradigm in which a person imagines a body movement without any physical action. By decoding the scalp-recorded electroencephalography (EEG) signals, BCIs can establish direct communication pathways to control external devices, offering significant potential in prosthetics, rehabilitation, and human–computer interaction. However, existing solutions remain difficult to deploy in practice. (i) Most employ independent, opaque models for each MI task. This fragmented methodology lacks a unified architectural foundation. Consequently, these models are trained in isolation and fail to learn robust representations from diverse datasets, which often results in modest performance. (ii) They primarily adopt fixed sensor deployment, whereas real-world setups vary in electrode number and placement, causing models trained on one configuration to fail under another. (iii) Performance degrades sharply under low-SNR conditions typical of consumer-grade EEG. Together, these limitations hinder the practical adoption of MI-based BCIs. To address these challenges, we present NeuroPath, a well-designed neural architecture for robust MI decoding. NeuroPath takes high-level inspiration from the brain’s signal pathway from cortex to scalp, utilizing a deep neural architecture with specialized modules for signal filtering, spatial representation learning, and feature classification, enabling unified decoding rather than task-specific black-box learning. To handle variations in electrode configurations, we introduce a spatially aware graph adapter that accommodates different electrode numbers and placements. To enhance robustness under low-SNR conditions, NeuroPath incorporates multimodal auxiliary training to refine EEG representations and stabilize performance on noisy, real-world data. Evaluations on three self-collected consumer-grade datasets and three public medical-grade datasets demonstrate that NeuroPath achieves superior performance.
- Open Access
- 25 February 2026
- Published by Association for Computing Machinery (ACM)
- p. 103-108
EarCalo: Earable-Based Energy Expenditure Estimation While Running
Accurately estimating energy expenditure (EE) is crucial for understanding exercise efficiency, managing fitness goals, and monitoring health conditions. Existing wearable systems either rely primarily on heart rate and motion sensors, resulting in unsatisfactory accuracy, or require bulky setups such as thermal cameras to improve performance by integrating more physiological information, which limits real-world applicability. We present EarCalo, an earable-based system that leverages in-ear audio sensing to estimate EE during running. The system extracts airflow-induced acoustic variations within the ear canal and employs a deep neural network to translate these subtle in-ear sound dynamics into EE estimates. The key insight is that in-ear acoustic signals can capture multiple physiological factors such as breathing and cardiovascular activity, while also reflecting motion-related cues like running intensity. These rich acoustic cues are closely related to EE and provide a unified sensing modality for estimation. We evaluated EarCalo on 21 participants running at varying speeds using a mixed-user setting. EarCalo achieved a mean absolute error (MAE) of 0.67 kcal/min, a mean absolute percentage error (MAPE) of 11.98%, and a Pearson correlation of 0.945, which achieves accuracy close to established physiological standards. This work represents an early step toward practical and personalized earable-based EE estimation in everyday settings.
- 25 February 2026
- Published by Association for Computing Machinery (ACM)
- p. 103-108
Open AccessEarCalo: Earable-Based Energy Expenditure Estimation While Running
Accurately estimating energy expenditure (EE) is crucial for understanding exercise efficiency, managing fitness goals, and monitoring health conditions. Existing wearable systems either rely primarily on heart rate and motion sensors, resulting in unsatisfactory accuracy, or require bulky setups such as thermal cameras to improve performance by integrating more physiological information, which limits real-world applicability. We present EarCalo, an earable-based system that leverages in-ear audio sensing to estimate EE during running. The system extracts airflow-induced acoustic variations within the ear canal and employs a deep neural network to translate these subtle in-ear sound dynamics into EE estimates. The key insight is that in-ear acoustic signals can capture multiple physiological factors such as breathing and cardiovascular activity, while also reflecting motion-related cues like running intensity. These rich acoustic cues are closely related to EE and provide a unified sensing modality for estimation. We evaluated EarCalo on 21 participants running at varying speeds using a mixed-user setting. EarCalo achieved a mean absolute error (MAE) of 0.67 kcal/min, a mean absolute percentage error (MAPE) of 11.98%, and a Pearson correlation of 0.945, which achieves accuracy close to established physiological standards. This work represents an early step toward practical and personalized earable-based EE estimation in everyday settings.
- Open Access
- 1 January 2026
- Published by Human Kinetics in Journal of Applied Biomechanics
- Vol. 42 (5) , 1-10
Ear-Worn Inertial Sensors Can Predict Gait Metrics and Reconstruct Vertical Ground Reaction Force Curves During Running
Ear-worn wearables (aka: earbuds, hearables, or earables) are commonly used by runners for entertainment, and many modern devices also include inertial sensors for user interaction. We propose harnessing the technology embedded in earbuds to capture fundamental aspects of running mechanics and make them available to the wider community of users, outside a lab setting. While other wearables such as insoles or ankle-/sacrum-mounted inertial measurement units have already been presented, ear-worn devices may have a better potential for adoption and therefore offer an optimal compromise between validity of running gait analysis and usability. Thirty healthy participants (18 males, 12 females) ran on an instrumented treadmill (54,000 gait cycles) and floor-mounted force plates (2800 gait cycles) at a variety of speeds. Building on the information brought about by the vibrations transmitted to, and motion of, the head, we devised a gait event detection algorithm and a regression model to predict vertical ground reaction force waveforms. The validation of outcomes against quantities from force plates shows an average mean absolute percentage error of 4.8% on temporal metrics and 9.0% on scalar ground reaction force derived metrics. Additionally, the model tracks the full vertical ground reaction force curve well, achieving an normalized root mean square error of 11.1% on unseen participants. Overall, we show comparable accuracy from an ear-worn consumer device in temporal and kinetic gait parameter estimation to specialist devices, paving the way for accessible running gait monitoring.
- 1 January 2026
- Published by Human Kinetics in Journal of Applied Biomechanics
- Vol. 42 (5) , 1-10
Open AccessEar-Worn Inertial Sensors Can Predict Gait Metrics and Reconstruct Vertical Ground Reaction Force Curves During Running
Ear-worn wearables (aka: earbuds, hearables, or earables) are commonly used by runners for entertainment, and many modern devices also include inertial sensors for user interaction. We propose harnessing the technology embedded in earbuds to capture fundamental aspects of running mechanics and make them available to the wider community of users, outside a lab setting. While other wearables such as insoles or ankle-/sacrum-mounted inertial measurement units have already been presented, ear-worn devices may have a better potential for adoption and therefore offer an optimal compromise between validity of running gait analysis and usability. Thirty healthy participants (18 males, 12 females) ran on an instrumented treadmill (54,000 gait cycles) and floor-mounted force plates (2800 gait cycles) at a variety of speeds. Building on the information brought about by the vibrations transmitted to, and motion of, the head, we devised a gait event detection algorithm and a regression model to predict vertical ground reaction force waveforms. The validation of outcomes against quantities from force plates shows an average mean absolute percentage error of 4.8% on temporal metrics and 9.0% on scalar ground reaction force derived metrics. Additionally, the model tracks the full vertical ground reaction force curve well, achieving an normalized root mean square error of 11.1% on unseen participants. Overall, we show comparable accuracy from an ear-worn consumer device in temporal and kinetic gait parameter estimation to specialist devices, paving the way for accessible running gait monitoring.
- Open Access
- 2 December 2025
- Published by Association for Computing Machinery (ACM) in Proceedings of the ACM on Interactive, Mobile, Wearable and Ubiquitous Technologies
- Vol. 9 (4) , 1-22
HearForce: Force Estimation for Manual Toothbrushing with Earables
Excessive tooth brushing force can contribute to oral health issues such as gum recession and bleeding. While some electric toothbrushes offer force feedback, manual toothbrushes remain widely used due to their affordability and accessibility but do not give this sort of information to users. Therefore, they often lack awareness of the force they apply while brushing. In this paper, we introduce HearForce, the first system to estimate tooth brushing force from manual toothbrushing using widely available earbuds. Unlike existing solutions that require significant modifications to toothbrushes, HearForce leverages in-ear microphones on commercial earbuds to capture bone-conducted toothbrushing sound propagating from the oral cavity to the ear canals. Our key insight is that variations in brushing force modulate these toothbrushing sounds due to the friction effect, allowing us to infer force levels through deep learning. However, individual habitual and anatomical differences introduce significant challenges for force estimation. To mitigate this, we propose a self-supervised representation learning network with a cross-attention mechanism to suppress user-dependent variability and a heuristic calibration strategy to adapt the model to different brushing habits. Through extensive evaluation, HearForce demonstrates force estimation capabilities with a Mean Absolute Error (MAE) of 37.3g in user-independent settings, corresponding to 11.4% of the typical force dynamic range. Our study makes the first step in the use of everyday earbuds for manual toothbrushing force monitoring, paving the way for accessible solutions to improve brushing habits for manual toothbrush users globally.
- 2 December 2025
- Published by Association for Computing Machinery (ACM) in Proceedings of the ACM on Interactive, Mobile, Wearable and Ubiquitous Technologies
- Vol. 9 (4) , 1-22
Open AccessHearForce: Force Estimation for Manual Toothbrushing with Earables
Excessive tooth brushing force can contribute to oral health issues such as gum recession and bleeding. While some electric toothbrushes offer force feedback, manual toothbrushes remain widely used due to their affordability and accessibility but do not give this sort of information to users. Therefore, they often lack awareness of the force they apply while brushing. In this paper, we introduce HearForce, the first system to estimate tooth brushing force from manual toothbrushing using widely available earbuds. Unlike existing solutions that require significant modifications to toothbrushes, HearForce leverages in-ear microphones on commercial earbuds to capture bone-conducted toothbrushing sound propagating from the oral cavity to the ear canals. Our key insight is that variations in brushing force modulate these toothbrushing sounds due to the friction effect, allowing us to infer force levels through deep learning. However, individual habitual and anatomical differences introduce significant challenges for force estimation. To mitigate this, we propose a self-supervised representation learning network with a cross-attention mechanism to suppress user-dependent variability and a heuristic calibration strategy to adapt the model to different brushing habits. Through extensive evaluation, HearForce demonstrates force estimation capabilities with a Mean Absolute Error (MAE) of 37.3g in user-independent settings, corresponding to 11.4% of the typical force dynamic range. Our study makes the first step in the use of everyday earbuds for manual toothbrushing force monitoring, paving the way for accessible solutions to improve brushing habits for manual toothbrush users globally.