Tao He
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Tao He has a total of 9 co-authors. Here are the top co-authors by number of shared publications:
- Qun Niu
School of Artificial Intelligence, Sun Yat-sen University, Zhuhai, Guangdong, China
4 shared publications - Hua-Bao Ling
Sun Yat-sen University, Guangzhou, Guangdong, China
2 shared publications Ning Liu
2 shared publications- Ning Liu
School of computer science and engineering, Sun Yat-Sen University, Guangzhou, Guangdong, China
2 shared publications - Suining He
Ubiquitous & Urban Computing Lab University of Connecticut Storrs CT USA
2 shared publications Gezhi Peng
School of Computer Science and Engineering and the Guangdong Key Laboratory of Information Security Technology, Sun Yat-sen University, Guangzhou, China
1 shared publication
Latest Publications
- Open Access
- 10 February 2026
- Published by Elsevier in SSRN Electronic Journal
Cold Start Multi-Source Collaborative Localization with Implicit Spatial Graph Encoding
Indoor localization is a core technology for Internet of Things applications, but existing indoor positioning methods tend to suffer from poor positioning accuracy due to sensor noise and the short duration of signals collected during the cold-start phase. To address these issues, we propose CC-Loc, a cold-start multi-source collaborative localization framework that leverages geomagnetic signals and Bluetooth Low Energy (BLE) for infrastructure-free, high-precision positioning. Specifically, geomagnetic signals capture location-specific fingerprint features, while BLE RSSI models inter-device relative spatial relationships without precise distance conversion. However, the limited observation length at system startup inevitably introduces ambiguity in initial position estimation. To alleviate this issue, we design a cold start enhancement module that generates reliable predicted positions from short signal sequences. Building upon this initialization, we further enhance an encoder–decoder architecture with adjacency matrix–based spatial encoding and a Spatial Pyramid Pooling (SPP) layer to adapt to dynamic collaborative graphs and extract topological information. Additionally, we suppress BLE noise by a multi-task learning framework jointly optimizes the primary task, coordinate prediction and the auxiliary task, relative distance prediction. Experimental results show that our method achieves an error of only 2.93m in a trial site of 1750m^2, even when the length of the geomagnetic sequence at the initial startup phase is only one second. It effectively addresses cold-start challenges and adapts to dynamic multi-device collaboration, making it suitable for consumer-grade indoor localization applications.
- 10 February 2026
- Published by Elsevier in SSRN Electronic Journal
Open AccessCold Start Multi-Source Collaborative Localization with Implicit Spatial Graph Encoding
Indoor localization is a core technology for Internet of Things applications, but existing indoor positioning methods tend to suffer from poor positioning accuracy due to sensor noise and the short duration of signals collected during the cold-start phase. To address these issues, we propose CC-Loc, a cold-start multi-source collaborative localization framework that leverages geomagnetic signals and Bluetooth Low Energy (BLE) for infrastructure-free, high-precision positioning. Specifically, geomagnetic signals capture location-specific fingerprint features, while BLE RSSI models inter-device relative spatial relationships without precise distance conversion. However, the limited observation length at system startup inevitably introduces ambiguity in initial position estimation. To alleviate this issue, we design a cold start enhancement module that generates reliable predicted positions from short signal sequences. Building upon this initialization, we further enhance an encoder–decoder architecture with adjacency matrix–based spatial encoding and a Spatial Pyramid Pooling (SPP) layer to adapt to dynamic collaborative graphs and extract topological information. Additionally, we suppress BLE noise by a multi-task learning framework jointly optimizes the primary task, coordinate prediction and the auxiliary task, relative distance prediction. Experimental results show that our method achieves an error of only 2.93m in a trial site of 1750m^2, even when the length of the geomagnetic sequence at the initial startup phase is only one second. It effectively addresses cold-start challenges and adapts to dynamic multi-device collaboration, making it suitable for consumer-grade indoor localization applications.
- 11 March 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Internet of Things Journal
- Vol. 12 (12) , 21740-21756
Spatio–Temporal Constrained Geomagnetic Indoor Localization With Arbitrary Walking Speed
This paper focuses on the accuracy and practicality of geomagnetic-based indoor localization in large-scale real-world sites. Previous studies often use continuous geomagnetic sequences for better discriminability but ignore that longer sequences, while more accurate, increase response times. Shorter sequences enhance practicality but sacrifice accuracy. Additionally, varying walking speeds can cause sequence variations along the same path, leading to ambiguity. To address these challenges, we propose a spatial-temporal constrained indoor localization model (STC-loc) that enables both accurate and practical localization using geomagnetic sequences collected at arbitrary walking speeds. Specifically, we first design a speed-oriented adaptive normalization module to tackle the walking speed heterogeneity problem. Then to balance accuracy and practicality, we initially employ shorter sequences as input and a self-attention hierarchical structure to efficiently extract the temporal features for initial location estimation. Subsequently, we further apply an Encoder-Decoder based refinement strategy, leveraging spatial contextual for continuous localization to ensure accuracy and robustness. Additionally, a slide-window based two-phase system architecture optimizes processing efficiency and scalability. Experiments on three large-scale real-world sites have demonstrated the superiority of the proposed STC-loc, reducing the localization error by more than 43% with shorter geomagnetic sequences.
- 11 March 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Internet of Things Journal
- Vol. 12 (12) , 21740-21756
Spatio–Temporal Constrained Geomagnetic Indoor Localization With Arbitrary Walking Speed
This paper focuses on the accuracy and practicality of geomagnetic-based indoor localization in large-scale real-world sites. Previous studies often use continuous geomagnetic sequences for better discriminability but ignore that longer sequences, while more accurate, increase response times. Shorter sequences enhance practicality but sacrifice accuracy. Additionally, varying walking speeds can cause sequence variations along the same path, leading to ambiguity. To address these challenges, we propose a spatial-temporal constrained indoor localization model (STC-loc) that enables both accurate and practical localization using geomagnetic sequences collected at arbitrary walking speeds. Specifically, we first design a speed-oriented adaptive normalization module to tackle the walking speed heterogeneity problem. Then to balance accuracy and practicality, we initially employ shorter sequences as input and a self-attention hierarchical structure to efficiently extract the temporal features for initial location estimation. Subsequently, we further apply an Encoder-Decoder based refinement strategy, leveraging spatial contextual for continuous localization to ensure accuracy and robustness. Additionally, a slide-window based two-phase system architecture optimizes processing efficiency and scalability. Experiments on three large-scale real-world sites have demonstrated the superiority of the proposed STC-loc, reducing the localization error by more than 43% with shorter geomagnetic sequences.
- Open Access
- 21 December 2022
- Published by Association for Computing Machinery (ACM) in Proceedings of the ACM on Interactive, Mobile, Wearable and Ubiquitous Technologies
- Vol. 6 (4) , 1-27
GC-Loc
Indoor localization techniques play a fundamental role in empowering plenty of indoor location-based services (LBS) and exhibit great social and commercial values. The widespread fingerprint-based indoor localization methods usually suffer from the low feature discriminability with discrete signal fingerprint or high time overhead for continuous signal fingerprint collection. To address this, we introduce the collaboration mechanism and propose a graph attention based collaborative indoor localization framework, termed GC-Loc, which provides another perspective for efficient indoor localization. GC-Loc utilizes multiple discrete signal fingerprints collected by several users as input for collaborative localization. Specifically, we first construct an adaptive graph representation to efficiently model the relationships among the collaborative fingerprints. Then taking state-of-the-art GAT model as basic unit, we design a deep network with the residual structure and the hierarchical attention mechanism to extract and aggregate the features from the constructed graph for collaborative localization. Finally, we further employ ensemble learning mechanism in GC-Loc and devise a location refinement strategy based on model consensus for enhancing the robustness of GC-Loc. We have conducted extensive experiments in three different trial sites, and the experimental results demonstrate the superiority of GC-Loc, outperforming the comparison schemes by a wide margin (reducing the mean localization error by more than 42%).
- 21 December 2022
- Published by Association for Computing Machinery (ACM) in Proceedings of the ACM on Interactive, Mobile, Wearable and Ubiquitous Technologies
- Vol. 6 (4) , 1-27
Open AccessGC-Loc
Indoor localization techniques play a fundamental role in empowering plenty of indoor location-based services (LBS) and exhibit great social and commercial values. The widespread fingerprint-based indoor localization methods usually suffer from the low feature discriminability with discrete signal fingerprint or high time overhead for continuous signal fingerprint collection. To address this, we introduce the collaboration mechanism and propose a graph attention based collaborative indoor localization framework, termed GC-Loc, which provides another perspective for efficient indoor localization. GC-Loc utilizes multiple discrete signal fingerprints collected by several users as input for collaborative localization. Specifically, we first construct an adaptive graph representation to efficiently model the relationships among the collaborative fingerprints. Then taking state-of-the-art GAT model as basic unit, we design a deep network with the residual structure and the hierarchical attention mechanism to extract and aggregate the features from the constructed graph for collaborative localization. Finally, we further employ ensemble learning mechanism in GC-Loc and devise a location refinement strategy based on model consensus for enhancing the robustness of GC-Loc. We have conducted extensive experiments in three different trial sites, and the experimental results demonstrate the superiority of GC-Loc, outperforming the comparison schemes by a wide margin (reducing the mean localization error by more than 42%).
- 20 September 2021
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Vehicular Technology
- Vol. 70 (11) , 11678-11694
Indoor Localization With Adaptive Signal Sequence Representations
Indoor location-based services (LBS) have exhibited large commercial and social values in smart cities, and urgent demands of which have spurred many localization techniques. Existing indoor localization approaches mostly rely on fingerprint techniques, leveraging either spatially discrete fingerprints or temporally consecutive ones for localization. However, these approaches often suffer from large errors or high time overhead in practice due to signal ambiguities or long input sequences. To overcome these drawbacks, this paper proposes a framework utilizing multiple adaptive representations of signal sequences for localization, where each representation indicates a corresponding signal structure with underlying location clues. As an example, the proposed approach takes geomagnetic signal sequences as input and infers location features from two intuitive representations, e.g., spatial and temporal ones. With adaptive signal representations, the proposed approach takes specifically optimized neural networks to extract corresponding location clues respectively and fuses them to generate more distinguishing features for more accurate localization. Furthermore, the ensemble learning mechanism is adopted in the approach and a weighted k-NN based location estimation algorithm is devised to enhance the robustness. Extensive experiments in three different trial sites demonstrate that the proposed approach outperforms state-of-the-art competing schemes by a wide margin, reducing mean localization error by more than 46%.
- 20 September 2021
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Vehicular Technology
- Vol. 70 (11) , 11678-11694
Indoor Localization With Adaptive Signal Sequence Representations
Indoor location-based services (LBS) have exhibited large commercial and social values in smart cities, and urgent demands of which have spurred many localization techniques. Existing indoor localization approaches mostly rely on fingerprint techniques, leveraging either spatially discrete fingerprints or temporally consecutive ones for localization. However, these approaches often suffer from large errors or high time overhead in practice due to signal ambiguities or long input sequences. To overcome these drawbacks, this paper proposes a framework utilizing multiple adaptive representations of signal sequences for localization, where each representation indicates a corresponding signal structure with underlying location clues. As an example, the proposed approach takes geomagnetic signal sequences as input and infers location features from two intuitive representations, e.g., spatial and temporal ones. With adaptive signal representations, the proposed approach takes specifically optimized neural networks to extract corresponding location clues respectively and fuses them to generate more distinguishing features for more accurate localization. Furthermore, the ensemble learning mechanism is adopted in the approach and a weighted k-NN based location estimation algorithm is devised to enhance the robustness. Extensive experiments in three different trial sites demonstrate that the proposed approach outperforms state-of-the-art competing schemes by a wide margin, reducing mean localization error by more than 46%.
- 1 December 2019
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 1-7
Indoor Localization with Spatial and Temporal Representations of Signal Sequences
Indoor localization has attracted considerable attention lately, due to its large commercial and social values in smart cities. The existing indoor localization approaches mostly rely on fingerprint techniques, and many of those leverage either spatially discrete fingerprints or temporally consecutive ones for localization, which either suffers from large errors due to signal ambiguities or high time overhead with long sequences. To achieve high accuracy with low computational cost, we propose ST-Loc, a deep neural network that extracts features from multiple representations of a single signal sequence for localization, where each representation indicates a corresponding signal structure with underlying feature correlations. Taking geomagnetism as an example, we infer location features from two different representations, e.g., spatial and temporal. In spatial representation, a signal sequence is converted to a signal heatmap, where each pixel corresponds to a spatial location and the value indicates fingerprint. Temporal representation, on the other hand, is a signal sequence with ordered readings, which provides temporal correlations. Using these different representations, we employ convolutional and recurrent networks to extract location features and fuse them to generate more distinguishing features for localization. We have conducted extensive experiments in two different trial sites, a narrow office area and a spacious food plaza. Our experimental results show that ST- Loc achieves more than 43% average localization error reduction compared with state-of-the-art competing schemes in both trial sites.
- 1 December 2019
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 1-7
Indoor Localization with Spatial and Temporal Representations of Signal Sequences
Indoor localization has attracted considerable attention lately, due to its large commercial and social values in smart cities. The existing indoor localization approaches mostly rely on fingerprint techniques, and many of those leverage either spatially discrete fingerprints or temporally consecutive ones for localization, which either suffers from large errors due to signal ambiguities or high time overhead with long sequences. To achieve high accuracy with low computational cost, we propose ST-Loc, a deep neural network that extracts features from multiple representations of a single signal sequence for localization, where each representation indicates a corresponding signal structure with underlying feature correlations. Taking geomagnetism as an example, we infer location features from two different representations, e.g., spatial and temporal. In spatial representation, a signal sequence is converted to a signal heatmap, where each pixel corresponds to a spatial location and the value indicates fingerprint. Temporal representation, on the other hand, is a signal sequence with ordered readings, which provides temporal correlations. Using these different representations, we employ convolutional and recurrent networks to extract location features and fuse them to generate more distinguishing features for localization. We have conducted extensive experiments in two different trial sites, a narrow office area and a spacious food plaza. Our experimental results show that ST- Loc achieves more than 43% average localization error reduction compared with state-of-the-art competing schemes in both trial sites.