M.A. Zuñiga Zamalloa
Please Note
79 records found
1
Artifact
Towards Device-Free Gaming with mmWave Radar
Developing millimeter-wave (mmWave) radar applications, particularly for human pose estimation (HPE), often requires substantial manual effort to build custom code for data access, visualization, and model training. In addition, the lack of shared dataset formats and module interfaces makes it difficult to reproduce results and to compare different systems under consistent preprocessing and evaluation settings. To address these challenges, we provide two toolkits, mwCore and mwPose3d, released alongside our paper [1] : mwCore [2] and mwPose3d [3]. Together, these artifacts provide a unified pipeline for processing radar point clouds, evaluating state-of-the-art tracking and HPE models, and deploying online applications via inference from streaming radar data, e.g., device-free gaming.
Don’t Pry Into My Personal Space
Exploiting mmWave Sensors to Provide Privacy for Areas With Glass Walls
Glass walls reduce illumination costs, enhance thermal comfort and improve social safety. However, they raise privacy concerns. People do not like passersby looking into personal areas, such as computer monitors. Our solution combines two upcoming technologies, mmWave sensors and transparent screens. The mmWave sensors detect passersby, and the transparent screen allows us to turn opaque only a small area instead of the entire wall. We build upon studies performing human pose estimation (HPE) from radar point clouds while addressing their two important shortcomings. They (i) focus on single static users, and (ii) estimate only the right pose, not the exact body location. We provide a solution that works for multiple mobile people, while simultaneously providing accurate location and pose estimation, enabling a robust identification of head coordinates. Our live evaluation shows that, for three passersby, dynamic opaque areas of 30 cm × 30 cm block the line of sight over 95% of the time. Furthermore, our method outperforms a competitive SoA alternative for HPE, reducing the RMSE error (worst-case error) for the head by a factor of 17. Our code and dataset will be publicly available to study human pose estimation in challenging mobile scenarios with multiple people.
Given the significant amount of time that people spend indoors, public spaces are reinventing themselves to deliver immersive experiences. One approach is to provide interactive gaming, such as the spaces created in big international airports like Singapore. The problem is that these gaming areas rely on either (i) wearable devices, which can easily spread germs, particularly in highly transited areas, such as those in airports, or (ii) cameras, which are increasingly raising privacy concerns and are being forbidden in some public areas. To provide an immersive gaming experience that is device-free and privacy-aware, we propose a system based on mmWave radar. In particular, our work provides three contributions. (1) For the first time, we compare various tracking and human pose estimation (HPE) models in the SoA under a unified framework. This approach allows us to identify the best methods in terms of position accuracy, latency, and smoothness-which are critical gaming metrics. (2) We build an integrated system with tracking and HPE capabilities. Our system includes the design of four games with different levels of complexity and a dataset collected specifically to train models for gaming applications. (3) Our evaluation, which includes a subjective users' survey and an objective comparison with a Kinect console, shows that today's mmWave is suitable for games where only the user's location is required or when coarse-grained gestures are needed. There are two key areas where mmWave research needs to improve to match camera-based consoles: sensors need a higher sampling rate, since a high frame rate is critical for seamless gaming, and HPE models need to master fine-grained gestures, in particular when arms end up in front of the body. All the code and datasets will be made available as a stepping stone for future work.
WiP Paper
EODL: Energy-Aware On-Device Training on Microcontrollers
Deep learning models can be trained on microcontrollers (MCUs) for privacy-sensitive applications, but such training is severely constrained by tiny SRAM capacity and limited battery energy. Existing on-device learning methods optimise memory usage by selectively freezing layers so that training fits within SRAM budgets, yet none explicitly account for training energy. A strategy that is feasible in memory may therefore still consume significantly more energy than necessary, shortening battery life on energyconstrained devices. We present eODL, an energy-aware on-device learning framework. eODL exploits two observations: (1) counts of multiplyaccumulate operations (MACs) predict training energy with high fidelity (R2 = 0.993), so energy can be estimated; and (2) energyaware layer-update selection can be solved efficiently with tabular dynamic programming. Using these, eODL selects which layers to update under a hard SRAM budget while balancing energy against accuracy. We implement eODL end to end on STM32 MCUs and evaluate it on MCUNetV2 fine-tuned on the Flowers-102 dataset. Under a 50 kB training-memory budget, eODL matches state-of-the-art accuracy while reducing measured on-device energy by 22.6% per epoch.
SolarML
Optimizing Sensing and Inference for Solar-Powered TinyML Platforms
HueLoc
Localization Through LEDs’ Hue Spectrum
Over the past decade, visible light positioning has become increasingly important for precise localization systems, yet its widespread adoption is limited due to the necessity of modifying existing lighting systems. This paper presents HueLoc, a novel method that bypasses this issue by using inherent features of light, such as the dominant colours in white LED lights, and employs affordable, energy-efficient hue sensors for location services. We propose that by extracting the power at dominant wavelengths of LEDs, these can be uniquely identified using a specifically designed signature. The unique signatures can be used by mobile objects for spatial awareness and further localization using the proposed regression-based learning approach. Our experiments demonstrate that HueLoc attains a location-mapping accuracy of 100% and achieves decimeter-level localization precision with a moving object in uncontrolled lighting conditions. Moreover, these unique signatures can be combined with other RF-based technologies to enhance their localization accuracy. As an example, this paper details the integration of Bluetooth features with light signatures using a three-stage incremental learning approach.
We propose a general framework to achieve reliable MIMO communications with passive-VLC. Our approach, which has a theoretical and empirical foundation, has three desirable properties: (i) does not assume orthogonality of the individual channels (overcomes co-channel interference), (ii) can exploit multiple properties of light (polarization and color); and (iii) is agnostic to LC parameters (which some studies rely on). Our results show that a transmitter with 9 LCs increases its capacity almost linearly up to 9 channels, attaining 6.8 kbps (750 bps per LC) using the simplest modulation method in the SoA. ...
We propose a general framework to achieve reliable MIMO communications with passive-VLC. Our approach, which has a theoretical and empirical foundation, has three desirable properties: (i) does not assume orthogonality of the individual channels (overcomes co-channel interference), (ii) can exploit multiple properties of light (polarization and color); and (iii) is agnostic to LC parameters (which some studies rely on). Our results show that a transmitter with 9 LCs increases its capacity almost linearly up to 9 channels, attaining 6.8 kbps (750 bps per LC) using the simplest modulation method in the SoA.
Edge-Light
Exploiting Luminescent Solar Concentrators for Ambient Light Communication
A recent advance in embedded Internet of Things (IoT) exploits ambient light for wireless communication. This new paradigm enables highly efficient links via simple light modulation, but the design space has a fundamental constraint: in most State of the Art (SoA) studies, the link can only follow the propagation direction of ambient light. Consider, for example, a swarm of drones and ground robots that want to communicate with sunlight. Drone-to-robot communication could be possible because sunlight travels downwards from the air (drone) to the ground (robot), allowing drones to modulate light to send information to robots beneath them. Robot-to-robot communication, however, is not possible because sunlight does not travel sideways (parallel to the ground). To allow ‘lateral communication’ with ambient light, we propose using Luminescent Solar Concentrators (LSC). These optical components receive ambient light on their surface and re-direct part of the spectra towards their edges. Considering this optical property of LSC, our work has three main contributions. First, we benchmark various optical properties of LSC to assess their performance for ambient light communication. Second, we combine LSC with liquid crystal (LC) shutters to form lateral links with ambient light. Third, we test our links indoors and outdoors with artificial and natural ambient light, by enhancing two robots with our LSC transceivers and showing that they can exchange basic commands and coordinate tasks by communicating only with sunlight.
SpectraLux
Towards Exploiting the Full Spectrum with Passive VLC
In recent years, the number of wireless applications has increased significantly, resulting in the radio bands becoming expensive and prone to interference. There is a new research area aiming at mitigating these issues by creating communication links using ambient light. This area, called passive-VLC, not only exploits the visible light frequencies, but does so with low-power transmitters. All the previous work in passive-VLC, however, forget about individual wavelength bands of light, and do not exploit its wide spectrum, reducing the potential channel capacity. In this paper, we propose a novel method to transmit and decode data, using liquid crystal cells that modulate and consider the full spectrum, and put it to the test by prototyping a multi-symbol communication link. The main contribution of our work is to show that passive-VLC can move from spectrum-agnostic to spectrum-aware modulation. We explore this new domain by making use of a novel type of receiver (i.e., a spectrometer) and uncovering the advantages and caveats of this spectrum-aware approach.
HueSense
Featuring LED Lights Through Hue Sensing
DroneVLC
Exploiting Drones and VLC to Gather Data from Batteryless Sensors
Cardiac patterns are being used to provide hard-to-forge biometric signatures in identification applications. However, this performance is obtained under controlled scenarios where cardiac signals maintain a relatively uniform pattern, facilitating the identification process. In this work, we analyze cardiac signals collected in more realistic (uncontrolled) scenarios and show that their high signal variability makes them harder to obtain stable and distinct features. When faced with these irregular signals, the state-of-the-art (SOTA) reduces its performance significantly. To solve these problems, we propose the CardioID framework1 with two novel properties. First, we design an adaptive method that achieves stable and distinct features by tailoring the filtering process according to each user’s heart rate. Second, we show that users can have multiple cardiac morphologies, offering us a bigger pool of cardiac signals compared to the SOTA. Considering three uncontrolled datasets, our evaluation shows two main insights. First, while using a PPG sensor with healthy individuals, the SOTA’s balanced accuracy (BAC) reduces from 90–95% to 75–80%, while our method maintains a BAC above 90%. Second, under more challenging conditions (using smartphone cameras or monitoring unhealthy individuals), the SOTA’s BAC reduces to values between 65–75%, and our method increases the BAC to values between 75–85%.
CardioID
Mitigating the Effects of Irregular Cardiac Signals for Biometric Identification
Cardiac patterns are being used to obtain hard-to-forge biometric signatures and have led to high accuracy in state-of-the-art (SoA) identification applications. However, this performance is obtained under controlled scenarios where cardiac signals maintain a relatively uniform pattern, facilitating the identification process. In this work, we analyze cardiac signals collected in more realistic (uncontrolled) scenarios and show that their high signal variability (i.e., ir-regularity) makes it harder to obtain stable and distinct user features. Furthermore, SoA usually fails to identify specific groups of users, rendering existing identification methods futile in uncontrolled scenarios. To solve these problems, we propose a framework with three novel properties. First, we design an adaptive method that achieves stable and distinct features by tailoring the filtering spectrum to each user. Second, we show that users can have multiple cardiac morpholo-gies, offering us a much bigger pool of cardiac signals and users compared to SoA. Third, we overcome other distortion effects present in authentication applications with a multi-cluster approach and the Mahalanobis distance. Our evaluation shows that the average balanced accuracy (BAC) of SoA drops from above 90% in controlled scenarios to 75% in uncontrolled ones, while our method maintains an average BAC above 90% in uncontrolled scenarios.