Real-time phonocardiogram analysis on embedded devices is a key enabler for scalable and accessible cardiovascular diagnostics, particularly considering portable systems designed for low-income countries. This work introduces a combination of compressive sensing and deep learning to build portable, efficient and effective diagnostic tools for widespread cardiac screenings. The proposed classification framework is tailored for deployment on ultra-low power STM32 microcontrollers equipped with the novel Neural-ART accelerator. Experimental evaluation on the CirCor Digiscope Phonocardiogram dataset demonstrates that even with a compression ratio exceeding 100×, a classification model can achieve up to 97.3% F1-score. A similar level of performance was obtained on the PhysioNet 2016 dataset, which was used to assess the robustness and generalization capability of the developed architectures. Compared to the state-of-the-art, our final edge solution achieves 94.3% accuracy, an inference time of 18.7 ms and an energy requirement of just 1.51 mJ per input window of 4,096 samples, confirming its suitability for real-time, energy-constrained medical applications.

The integration of compressive sensing with realtime embedded systems opens new possibilities for efficient, low-power biomedical signal acquisition. This paper presents a custom hardware platform based on the RP2350 microcontroller, tailored for synchronized multi-modal biomedical monitoring. The system is capable of capturing cardiopulmonary sounds, along with biopotential signals such as phonocardiography (PCG), electrocardiography (ECG) and electromyography (EMG), photoplethysmography (PPG), and inertial measurement unit (IMU) data for posture recognition. To ensure sampleaccurate synchronization, a Sub-1GHz radio system is used across multiple nodes. Wi-Fi and Bluetooth connectivity enable centralized data aggregation. Experimental results demonstrate the achieved decrease in power consumption when using compressive sensing, efficient multi-node synchronization, and scalability for wireless biomedical monitoring applications. The compact form factor and low-cost design make it suitable for various medical applications, including remote healthcare and long-term monitoring.

In music performance research, conservatory pianists have been a subject of interest. However, previous studies have often relied on data captured from unsynchronised devices, leaving a significant portion of potential data unexplored. This paper introduces a synchronisation method for a multimodal sensor setup to address this gap. The primary focus of this paper is the implementation of a synchronization mechanism. The proposed technique is a cheap and robust way to synchronise the required multimodal setup. The insights derived from this research can improve future studies and methodologies in musical performance research and are more generically applicable towards any multimodal sensor setup.

Digital auscultation is a well-known method for assessing lung sounds, but remains a subjective process in typical practice, relying on the human interpretation. Several methods have been presented for detecting or analyzing crackles but are limited in their real-world application because few have been integrated into comprehensive systems or validated on non-ideal data. This work details a complete signal analysis methodology for analyzing crackles in challenging recordings. The procedure comprises five sequential processing blocks: (1) motion artifact detection, (2) deep learning denoising network, (3) respiratory cycle segmentation, (4) separation of discontinuous adventitious sounds from vesicular sounds, and (5) crackle peak detection. This system uses a collection of new methods and robustness-focused improvements on previous methods to analyze respiratory cycles and crackles therein. To validate the accuracy, the system is tested on a database of 1000 simulated lung sounds with varying levels of motion artifacts, ambient noise, cycle lengths and crackle intensities, in which ground truths are exactly known. The system performs with average F-score of 91.07% for detecting motion artifacts and 94.43% for respiratory cycle extraction, and an overall F-score of 94.08% for detecting the locations of individual crackles. The process also successfully detects healthy recordings. Preliminary validation is also presented on a small set of 20 patient recordings, for which the system performs comparably. These methods provide quantifiable analysis of respiratory sounds to enable clinicians to distinguish between types of crackles, their timing within the respiratory cycle, and the level of occurrence. Crackles are one of the most common abnormal lung sounds, presenting in multiple cardiorespiratory diseases. These features will contribute to a better understanding of disease severity and progression in an objective, simple and non-invasive way.

Biologically inspired algorithms for simultaneous localization and mapping (SLAM) such as RatSLAM have been shown to yield effective and robust robot navigation in both indoor and outdoor environments. One drawback however is the sensitivity to perceptual aliasing due to the template matching of low-dimensional sensory templates. In this paper, we propose an unsupervised representation learning method that yields low-dimensional latent state descriptors that can be used for RatSLAM. Our method is sensor agnostic and can be applied to any sensor modality, as we illustrate for camera images, radar range-doppler maps and lidar scans. We also show how combining multiple sensors can increase the robustness, by reducing the number of false matches. We evaluate on a dataset captured with a mobile robot navigating in a warehouse-like environment, moving through different aisles with similar appearance, making it hard for the SLAM algorithms to disambiguate locations.

Navigating spatially varied and dynamic environments is one of the key tasks for autonomous agents. In this paper we present a novel method of navigating a mobile platform with one or multiple 3D-sonar sensors. Moving a mobile platform and subsequently any 3D-sonar sensor on it, will create signature variations over time of the echoed reflections in the sensor readings. An approach is presented to create a predictive model of these signature variations for any motion type. Furthermore, the model is adaptive and works for any position and orientation of one or multiple sonar sensors on a mobile platform. We propose to use this adaptive model and fuse all sensory readings to create a layered control system allowing a mobile platform to perform a set of primitive motions such as collision avoidance, obstacle avoidance, wall following and corridor following behaviours to navigate an environment with dynamically moving objects within it. This paper describes the underlying theoretical base of the entire navigation model and validates it in a simulated environment with results that shows the system is stable and delivers expected behaviour for several tested spatial configurations of one or multiple sonar sensors that can complete an autonomous navigation task.

Bat-pollinated flowers have to attract their pollinators in absence of light and therefore some species developed specialized echoic floral parts. These parts are usually concave shaped and act like acoustic retroreflectors making the flowers acoustically conspicuous to the bats. Acoustic plant specializations only have been described for two bat-pollinated species in the Neotropics and one other bat-dependent plant in South East Asia. However, it remains unclear whether other bat-pollinated plant species also show acoustic adaptations. Moreover, acoustic traits have never been compared between bat-pollinated flowers and flowers belonging to other pollination syndromes. To investigate acoustic traits of bat-pollinated flowers we recorded a dataset of 32320 flower echoes, collected from 168 individual flowers belonging to 12 different species. 6 of these species were pollinated by bats and 6 species were pollinated by insects or hummingbirds. We analyzed the spectral target strength of the flowers and trained a convolutional neural network (CNN) on the spectrograms of the flower echoes. We found that bat-pollinated flowers have a significantly higher echo target strength, independent of their size, and differ in their morphology, specifically in the lower variance of their morphological features. We found that a good classification accuracy by our CNN (up to 84%) can be achieved with only one echo/spectrogram to classify the 12 different plant species, both bat-pollinated and otherwise, with bat-pollinated flowers being easier to classify. The higher classification performance of bat-pollinated flowers can be explained by the lower variance of their morphology.

In this paper, we introduce a comprehensive angle of arrival (AoA) estimation solution for the long range (LoRa) communication network. Termed the LoRa array (LoRay), the proposed system constitutes hardware and software solutions to estimate the AoA of the received signals in real life urban environments. The hardware solution is based on converting multiple individual software defined radios (SDR) into a single SDR that consists of multiple RF-channels. The proposed hardware is cost effective, flexible and generic. The software solution, on the other hand, utilizes the space alternating generalized expectation-maximization (SAGE) algorithm to estimate the AoA of highly correlated received signals. The proposed software exploits few samples of the received signal to estimate the AoA of the direct and reflected paths in an intensive multipath environment. The LoRay system has been validated in outdoor urban environments. The experimental results show that the proposed system provides stable and accurate AoA estimates for both the line-of-sight (LoS) and the non-line-of-sight (NLoS) conditions. The AoA of 80% of the received signals have been estimated within an estimation error below 2° and 10° for the LoS and the NLoS locations, respectively.

State-of-the-art autonomous vehicles mainly rely on optical sensors to perceive their environment. However, the performance of these sensors worsens dramatically in environments where airborne particles are present. Sonar sensors rely on acoustic waves which are able to pass through these distortions. The data gathered from sonar could complement the distorted data of the optical sensors in these earlier mentioned environments. In this paper we will discuss the newest 3D in-air sonar sensor developed by CoSys-Lab: the micro Real Time Imaging Sonar (μRTIS). It is the smaller version of the previously developed embedded Real Time Imaging Sonar (eRTIS) and consists of a uniform rectangular array of 5 by 6 microphones compared to the pseudo-random 32 microphone array based on Poisson Disc-Sampling used with the eRTIS. In this paper we will discuss the used hardware of the μRTIS, followed by the implemented processing algorithms to form acoustic images from reflected ultrasonic data. Furthermore, we will discuss the spatial resolution that can be obtained with the presented hardware architecture using different beamforming algorithms. Finally, we will validate the real-life performance, comparing the acoustic images of the μRTIS when using Delay-And-Sum or MUltiple SIgnal Classification to solve the different directions of arrival.

This paper presents a proof-of-concept for a localization and mapping system for magnetic crawlers performing inspection tasks on structures made of large metal plates. By relying on ultrasonic guided waves reflected from the plate edges, we show that it is possible to recover the plate geometry and robot trajectory to a precision comparable to the signal wavelength. The approach is tested using real acoustic signals acquired on metal plates using lawn-mower paths and random-walks. To the contrary of related works, this paper focuses on the practical details of the localization and mapping algorithm.

Many popular advanced sonar systems provide accurate and reliable measurements containing crucial info needed by robotic applications such as range, bearing and reflection strength of the objects in the field of view. While these sensor systems provide these crucial pieces of information accurately, they are often limited by a lack of processing power and/or size which leads to them needing an external computing device to process all the information generated by the microphone array on the sensor. In this paper we present two versions of a novel fully embedded 3D sonar sensor which have different sensing architectures which enable 3D perception for robotic application in harsh conditions using ultrasound at low cost. Experimental results taken from an office environment will show the 3D localization capabilities and performance of the sensor, showing the sensor has a large field-of-view (FoV) with accurate 3D localization combined with real-time capabilities.

In this paper we present a flexible low-cost sonar sensor platform that can be used for a wide range of biomimetic sonar experiments and autonomous sonar navigation targeted at robotics applications. The navigation abilities of bats using ultrasound (sonar) in unknown cluttered environments are very effective and can be distilled into a sensor architecture and accompanying control methodology that lends itself to be implemented on cost efficient hardware. The sensor architecture and processing methodology of this sensing platform mimics that of bats. In this paper we specifically focused on the common big-eared bat (Micronycteris microtis) although this could be transferred to other bat species or even other echolocating animals since the experimental platform was designed for flexibility. Using this platform we were able to implement a control system using a subsumption architecture that features different behavior patterns based solely on the sonar sensor as a source of exteroceptive information. In order to validate the combination of our autonomous navigation control system and our developed sonar sensor platform, the hardware was mounted on the P3DX robotics platform that was introduced in an unknown testing environment and have it drive autonomously. These experiments were used to validate our assumption of the efficacy of these relatively simple biomimetic control mechanisms and thus alleviating the need for expensive sensing platforms for certain robotics applications.

Obtaining accurate data about the environment in which a robot is located is a crucial matter when it comes to autonomous navigation and other robotic applications. A popular method of acquiring this information is to use sonar-rings, where a robot is fitted with multiple simple ultrasound transducers pointed in the directions where an object can appear. However, in a time where accurate 3D data is gaining importance, other sensing modalities are becoming more popular because of the ability to measure dense 3D point clouds. In these point clouds not only the horizontal plane is measured, but objects in the elevation planes can also be registered, which can be very interesting and makes applications such as 3D SLAM or object recognition possible. In this paper we present a way to extract complex 3D point cloud data from the entire surrounding sphere using multiple interconnected eRTIS sensors. These advanced imaging sonar sensors offer the flexibility of the popular sonar-ring in combination with the benefits of some of the competing sensing modalities. The setup presented here uses less sonar sensors (and thus less external hardware) while obtaining more information from the complete frontal hemispheres of each individual sensor. This setup is discussed, along with the issues that arise when using complex imaging sonar sensors in a network, and is tested in an indoor and outdoor environment. At the end of this paper is a discussion of the obtained results.

Most objects and vegetation making up the habitats of echolocating bats return a multitude of overlapping echoes. Recent evidence suggests that the limited temporal and spatial resolution of bio-sonar prevents bats from separately perceiving the objects giving rise to these overlapping echoes. Therefore, bats often operate under conditions where their ability to localize obstacles is severely limited. Nevertheless, bats excel at avoiding complex obstacles. In this paper, we present a robotic model of bat obstacle avoidance using interaural level differences and distance to the nearest obstacle as the minimal set of cues. In contrast to previous robotic models of bats, the current robot does not attempt to localize obstacles. We evaluate two obstacle avoidance strategies. First, the Fixed Head Strategy keeps the acoustic gaze direction aligned with the direction of flight. Second, the Delayed Linear Adaptive Law (DLAL) Strategy uses acoustic gaze scanning, as observed in hunting bats. Acoustic gaze scanning has been suggested to aid the bat in hunting for prey. Here, we evaluate its adaptive value for obstacle avoidance when obstacles can not be localized. The robot's obstacle avoidance performance is assessed in two environments mimicking (highly cluttered) experimental setups commonly used in behavioral experiments: a rectangular arena containing multiple complex cylindrical reflecting surfaces and a corridor lined with complex reflecting surfaces. The results indicate that distance to the nearest object and interaural level differences allows steering the robot clear of obstacles in environments that return non-localizable echoes. Furthermore, we found that using acoustic gaze scanning reduced performance, suggesting that gaze scanning might not be beneficial under conditions where the animal has limited access to angular information, which is in line with behavioral evidence.

This paper presents a novel biomimetic radar sensor for autonomous navigation. To accomplish this, we have drawn inspiration from the sensory mechanisms present in an echolocating mammal, the common big-eared bat (Micronycteris microtis). We demonstrate the correspondence in both the hardware, system model, and signal processing. To validate the performance of the sensor, we have developed a complementary control system based on subsumption architecture, which allows the system to autonomously navigate unknown environments. This architecture consist of separate behaviors with different levels of complexity, which are combined to produce the overall functionality of the system. We describe each behavior separately and examine their performance in real-world navigation experiments. For this purpose, the system is placed in two distinct office environments with the goal of achieving smooth and stable trajectories. Here, we can observe noticeable improvements when employing high-level behaviors. Furthermore, we utilize the data collected during the navigation experiments to perform simultaneous localization and mapping, using an algorithm developed in our earlier work. These results show a substantial improvement over the odometry. We attribute this to the fact that the system traverses stable and repetitive paths, which facilitates place recognition.

In this paper we will propose a method to evaluate the performance of a certain ultra-wide band fixed anchor configuration in complex indoor environments by making use of the mutual information as the performance metric. Furthermore we will introduce an incremental algorithm that will determine an optimized anchor configuration for complex indoor environments. By making use of heuristics we are able to ensure that the time required to complete the algorithm is feasible on commercial grade computers, even for large-scale floor plans.

In this paper we describe a system for passive acoustic sound source localization in 3D using a distributed microphone array. By distributing the microphones into small-scale and large-scale arrays we can exploit array processing algorithms that use near-field and far-field properties of the sound source with respect to the array. The microphones on the small-scale device are sampled using a single, simultaneously sampling analog to digital converter which results in time synchronization between these microphone channels up to a small fraction of the signal's phase. The large-scale array consists of multiple small-scale devices and can be deployed over a much larger capture volume. We create the distributed microphone arrays in a manner that is scale-free with regard to the amount of microphones in the system, the capture volume, the type of sound and the type of sensors/microphones. In this paper we describe the hardware topology, the localization algorithms and the obtained results.

Designing model-based physical systems has growing demand in consequence of increasing system complexity. In particular, observers/estimators are extensively used for the applications requiring state or disturbance estimation. Designing and deploying such numerically intensive physical systems onto embedded targets is a challenging task that requires codesign among various stakeholders from different technical backgrounds.The most important challenge is to obtain a numeric behavior of the estimator from an embedded target, that is able to represent the physical system states/disturbance with an acceptable error margin. Moreover, this error margin needs to be decided by the stakeholders, which makes the overall embedded deployment a co-design problem. The main contribution of this paper is to investigate the cause of the estimation error of an estimator that is deployed to embedded targets. This error is studied in the form of precision loss in addition to the error originating in the decreasing estimator measurement frequency for the embedded targets. We propose Assume-Guarantee (A/G) contracts to reconcile the viewpoints of the stakeholders, who reside at different abstraction levels. The feasibility of the proposed physical system deployment method is presented by utilizing a model-based virtual sensor estimator deployment for embedded targets as a case study.

The industrial demands for accurate localization systems have been rapidly increasing after the introduction of the Internet of Things (IoT) concept. Self localization and tracking transmitting sources are considered essential parts of IoT applications. In this paper we studied the possibility of applying angle of arrival (AoA) estimations to localize an IoT transceiver device in an indoor environment. Furthermore, we propose an adaptive probabilistic model which works on top of the AoA estimation technique to improve the localization accuracy. The experimental results show the potential of using AoA-based localization for indoor environments. The results furthermore show that the proposed adaptive probabilistic model outperforms the traditional static probabilistic model in terms of the localization accuracy and the stability of the position estimate.

Pose estimation systems have seen some big developments in the last two decades due to technological advances and a greater need for these systems. The industry that thrives and simultaneously popularizes these developments is the entertainment industry. One of the latest developments in home entertainment are Virtual Reality systems that lets users have an immersive experience when playing video games. In order to fully achieve this immersive experience, objects in the ‘real world’ can be used to manipulate objects in the virtual world. Therefore the ‘real world’ objects need to be located in the user's environment to a great extent. One of these virtual reality kits utilizes an optical solution that incorporates dual-axis rotating laser sweeps on the transmitter side and photodiodes on the receiver side to achieve the positioning of these objects. Since the transmitters are stand-alone systems that can be purchased at a relatively low price, we propose to use this hardware in combination with custom low-cost receiver hardware to achieve an affordable yet precise and accurate six degrees-of-freedom human body pose estimation system.

This paper presents a novel method for using an ultra-wideband (UWB), super high frequency (SHF) pulse-echo radar sensor as a biomimetic sensing mechanism to successfully solve the Simultaneous Localization and Mapping (SLAM) problem. Due to recent advances in consumer radar technology it has become possible to sample the received echo signals well above their Nyquist frequency. This Nyquist-conform sampling permits the conversion of the signal waveforms into spectrograms which contain spatiospectral cues caused by the interactions of the echoes with features of the environment and the antenna's radiation pattern. Such spectrograms can therefore serve as distinct labels for individual locations, allowing for the identification and recognition of these locations. By adapting an existing acoustic SLAM system (BatSLAM) we have developed a system that demonstrates the feasibility of our proposed method; the results validate the potential of using pulse-echo radar as an exteroceptive sensory modality in topological SLAM systems.

This paper describes an alternative approach to giving people with limited hand and arm movement the ability to select objects on a computing device using head movement (head mouse). To allow for better filtering of unintentional head movement, and to allow for a faster update rate of the head mouse, the full six-DoF pose of the head is estimated using a low-cost camera and IR markers, a 3-axis accelerometer, and a 3-axis magnetometer. The pose estimation problem was cast in a probabilistic fashion in which information from the different sensors is fused into a single a-posteriori distribution for the sensor pose. Simulations were run to analyze the influence of the proposed sensor fusion algorithm on the stability and accuracy of the pose estimation, and thus on the ability to point a mouse cursor to a specified location on a screen. Experiments were then performed validating the results from the simulations on real sensors. The proposed algorithm was shown to give more accurate and stable results than using only a camera to estimate the six-DoF pose.

This paper introduces and validates the performance of an alternative input device for people with limited hand/arm movement and control. A summary is provided of the current state of the art in alternative input devices. Based on this, a low-cost solution is proposed that allows for 1) low latency and high operating speed and accuracy, 2) use from different viewing angles without recalibration, and 3) the ability to seamlessly control multiple devices. A prototype of this system was built and tested to determine the accuracy of the system (usingapan-tilt system), and to analyze the performance of the system compared to the state of the art (through user tests, based on the ISO 9241-411 test). The proposed system allows for great accuracy (σX= 0.28 px, σY= 0.29 px, σXY= 0.02 px), a decent performance compared to the state of the art (throughput = 1.52 bits/s, error rate = 31%), and good results in the ISO 9241-411 independent rating scale. The results from the performed experiments show that the proposed system leads great promise in real-world applications. A low-cost head-mounted camera could be used as an alternative human interface device for people with limited hand/arm movement and control, allowing them to participate in the ongoing trend of computing devices gaining importance in our everyday activities.

Throughout the last decades, human motion capture systems have become an important tool for various sectors. Besides the entertainment sector, which is probably the most known sector to use these systems through popular movies and video games, the medical sector has adopted this technology as an analysis tool. These clinical analyses measure the human body posture and movement for various purposes including rehabilitation and sports, and require high accuracy and precision. Although there is a variety of products that offer these capabilities, there is not yet an affordable system that meets the requirements for medical implementations. We therefore propose using an ultrasonic six degrees-of-freedom sensor system which combines a distributed array of ultrasonic emitters and a small array of ultrasonic receivers. In order to support low-cost human pose estimation, while retaining the possibility for re-use and expandability, the need for a flexible embedded hardware platform naturally emerged. The proposed hardware design supports the requirements of a high accuracy and high precision pose estimation system while keeping its overall system cost low.

Passive acoustic localization is an important technique in a wide variety of monitoring applications, ranging from healthcare over biological survey to structural health monitoring of buildings. If the subject of interest emits a recognizable sound and is picked up by a distributed array of microphones it is possible to measure the difference in arrival times and reconstruct the source positions. This requires the microphones to be synchronized up to a fraction of the expected time differences of arrivals in order for the system to be able to produce accurate location estimates. In this paper we take a closer look at the techniques required to synchronize a network of sensors, connected through a low-power RF-communication link, in a distributed manner, ie. without the presence of a master node. The absence of a master node makes the network more robust against the failure of a single node. We took inspiration from the synchronization technique observed in some species of fireflies. A synchronized pseudo random number is embedded in the captured data as a marker to re-align the data-streams in time in a post-processing stage.

In this paper, we present a solution to the Simultaneous Localization and Mapping problem by combining a novel 3D in-air sonar sensor with techniques for estimating the egomotion of a mobile platform (called acoustic flow odometry) and a bio-inspired mapping module (called BatSLAM). This combination eliminates the need for odometric information originating from the robot's motor controller, enabling applications where that information is difficult to obtain, e.g. many electric wheelchairs. The proposed method exploits the programmable spatial sampling capabilities of the 3D sonar system to derive from the same set of received echo signals both an accurate representation of reflectors in the horizontal plane (2D energyscape) and a more coarse representation of reflectors in the frontal hemisphere (3D energyyscape). Using a mapping experiment we demonstrate that the motion estimates originating from the acoustic flow module combined with the coarse frontal hemisphere data provide sufficient information for the mapping module to reconstruct the robot's trajectory.

In this paper we explore the use of 3D acoustic flow fields to steer robot motion. We derive the 3D velocity fields set up by linear and rotational robot motions and explain how they can be sampled directly by a sonar array-sensor that we developed recently. The resulting acoustic flow field patterns are then shown to contain all the information necessary for controlling obstacle avoidance and corridor following behavior. Experimental data collected by the sonar system mounted on a wheelchair that was driven in an office environment are presented to validate the theoretically predicted acoustic flow patterns.

Array beamforming techniques allow for the generation of 3-D spatial filters which can be used to localize objects in a large field of view (FOV) without the need for mechanical scanning. By combining broadband beamforming with a sparse, random array of receivers, we have constructed a low-cost, yet powerful, in-air sonar system, which is suited for a wide range of robotic applications. Experimental results in unmodified office environments show the performance of the sonar sensor. In particular, we document the sensor's capacity to produce 3-D location measurements in the presence of multiple highly overlapping echoes. We show how this capability makes possible the combination of a wide FOV with accurate 3-D localization, allowing the sensor to operate under real-time constraints in realistic environments. To demonstrate the use of this sensor, we describe an odometry application that estimates egomotion of a mobile robot using acoustic flow.

We describe how, by combining broadband beamforming with a sparse, random array of receivers, we have constructed a low-cost, yet powerful, in-air sonar system which is suited for a wide range of robotic applications. In particular, we show how simple array beamforming techniques allow for the generation of 3D spatial filters that can be used to accurately localize objects in a large field of view without the need for mechanical scanning. Experimental validation of the sensor's capacity to produce 3D location measurements in the presence of multiple highly overlapping echoes is presented.