System and method for proactive cardiorespiratory health assessment using capacitive sensing of electric field perturbations

Abstract

Disclosed are devices, systems, and methods for non-invasive cardiorespiratory monitoring and assessment. A capacitive sensor detects fluctuations in an electric displacement field around a subject, caused by changes in accumulated static charge modulated by cardiorespiratory activity (e.g., heart function and breathing dynamics). A processor determines cardiorespiratory parameters from sensor data representing these fluctuations. Systems may include a central processing unit for analysis, comparison to thresholds, trend identification, and generating alerts. Proactive assessment systems utilize artificial intelligence (AI) models, potentially integrating supervised and reinforcement learning, to analyse historical and real-time data, predict future cardiorespiratory health conditions, and generate timely alerts. The method involves positioning the sensor(s), detecting field fluctuations, generating data, and processing it, potentially using machine learning or AI. This approach senses static charge fluctuations, differing from prior art methods focused on volume or dielectric changes.

Description

SYSTEM AND METHOD FOR PROACTIVE CARDIORESPIRATORY HEALTH ASSESSMENT USING CAPACITIVE SENSING OF ELECTRIC FIELD PERTURBATIONSTECHNICAL FIELD

[0001] The present disclosure relates generally to the field of non-invasive physiological monitoring. More specifically, it pertains to capacitive sensing devices, systems, and methods for detecting cardiorespiratory activity in human or animal subjects by sensing perturbations or fluctuations in the electric displacement field surrounding the subject, wherein these fluctuations are linked to changes in static charge accumulated on the body and modulated by physiological functions like heart activity and breathing dynamics. The disclosure further relates to the processing of signals obtained via such sensing for cardiorespiratory monitoring and proactive health assessment, potentially utilising artificial intelligence.BACKGROUND

[0002] The state of human health is determined by many interdependent physiological parameters. Not all of them are equally informative and important. In addition, not all of these parameters can be easily and accurately controlled, since their measurement requires special conditions and expensive medical equipment and materials. When designing a general medical sensing device, it is necessary to evaluate not only the importance of the parameters being measured, but also the methods for their measurement and the possibility of implementation in practical systems. Medical research has proven that the most important parameters are those that determine the functioning of the heart and the respiratory or pulmonary system.

[0003] The cardiovascular and pulmonary systems , often referred to collectively as the cardiorespiratory system, are inherently multidimensional, closed-loop systems. The basic physiological parameters of these systems are closely interrelated, mediated by various mechanical and neural mechanisms, which can also be assessed by monitoring brain activity. The autonomic nervous system controls various parts of this system through continuous neural modulation, causing variables to differ slightly from their average values. Breathing as an external stimulator can significantly modulate hemodynamic parameters.The connection and interaction between brain activity, circulation and respiration are very close, highlighting the need for simultaneous monitoring and assessment of these systems in clinicalwork.

[0004] A number of different types of devices are currently available for monitoring people in a non-invasive manner. For example, a patient's heart function can be monitored using electrodes that need to be attached to the patient's skin. Although such equipment is non-invasive, it is nevertheless inconvenient for the patient, who is connected to a network of cables and wired sensors. In addition, such equipment is expensive, so its use is limited to hospitals, laboratories and other healthcare settings where both the cost and the discomfort to patients can be justified.

[0005] In general, non-invasive and point-of-care diagnostics of a person is desirable. To enable regular monitoring of people in their usual environment, such as home or office, the equipment must be non-invasive and easy to use. For example, such monitoring can be very useful as part of the overall health maintenance of a patient and can be used to detect deterioration in the patient's physiological condition before corresponding deterioration becomes noticeable.

[0006] Typical equipment used in hospitals and laboratories is capable of monitoring at least one physiological parameter of a patient without requiring the patient to perform any complex actions and / or operate complex devices. However, as mentioned above, it would be highly preferable for the equipment to be included in the normal daily life of the user or patient, since the need for any additional or special actions on the part of the patient may result in reduced compliance. In addition, the equipment must be reliable but inexpensive, such as a small bracelet for attaching a physiological sensor to the user's wrist. There are currently several different types of such bracelets available, which are mentioned below, most of which are designed to be used as stand-alone devices to provide information about the user's own physical condition, mainly heart rate and blood pressure.

[0007] A pulse wave is the change in the volume of a blood vessel that occurs when the heart pumps blood, and a detector that monitors this volume change is called a pulsesensor. There are four main ways to measure heart rate: electrocardiogram, photoelectric pulse wave, blood pressure measurement, and phonocardiography.

[0008] When assessing a person's health, it is always advisable to record physiological indicators related to the work of the heart, since the work of the heart is associated with physiological functions that are vital for the human body. The main physiological parameters associated with the work of the heart are heart rate (pulse) and blood pressure. The electrocardiogram (ECG) is one of the most representative physiological characteristics of the heart. Electrocardiography is the process of recording the electrical activity of the heart over a period of time using electrodes placed on the patient's body. These electrodes detect tiny electrical changes in the skin resulting from the polarisation and depolarisation of the heart muscle during each heartbeat, and with their help and / or another device such as an electrode reader, a recording is generated.

[0009] Conventional electrocardiographs use multiple (8-12) electrodes to measure the electrical activity of the heart. Each electrode is placed on the patient in a specific location with a specific tolerance. With these electrodes, the overall magnitude of the heart's electrical potential is typically measured from twelve different angles ("leads") and recorded over a specified period of time (usually 10 seconds). Thus, the overall magnitude and direction of the electrical dipole of the heart are recorded at each moment of the cardiac cycle. The resulting graph of voltage versus time obtained with this non-invasive medical procedure is called an electrocardiogram. Various traditional electrocardiographs include 3, 5, 15, 16, etc. leads.

[0010] During each heartbeat in the healthy heart, there is an orderly progression of depolarisation and polarisation that begins with the pacemaker cells in the sinoatrial node, spreads through the atrium, passes through the atrioventricular node down into the bundle of HIS and into the Purkinje fibres spreading down and to the left throughout the ventricles. This orderly pattern of depolarisation and polarisation gives rise to the ECG tracing.

[0011] Medical information obtained with ECG is indispensable for directly determining the function of the heart and many other vital parameters of the body, such as heart disease, atrial fibrillation, cholesterol blockade, heart attack prediction, hypertension and manyothers. An ECG conveys a large amount of information about the structure of the heart and the function of its electrical conduction system. Among other things, an ECG can be used to measure the rate and rhythm of the heart, the size and position of the heart's chambers, the presence of any damage to the heart's muscle cells or conduction system, the effects of cardiac drugs, and the function of implanted pacemakers.

[0012] The ability to detect ECGs frequently in daily life is extremely important and invaluable for predictive medical care in today's society, where signals of heart failure can be detected as early as possible. However, the ECG technique is cumbersome, minimally invasive, requires the installation of several electrodes on the body and, therefore, cannot be used at a single point of the patient's body and requires an outpatient presence. Recently, pulse watches have come on the market that measure the heart rate at one point on the wrist of a user. The pulse watches use photoplethysmography (PPG) to continuously measure hemodynamic blood waves and heart rate.

[0013] Pulse sensors using the photoelectric pulse wave method are classified into two types depending on the measurement method: transmission and reflection. Transmission types measure pulse waves by emitting red or infrared light from the body surface and detecting the change in blood flow during heart beats as a change in the amount of light transmitted through the body. This method is limited to areas where light can easily penetrate, such as the fingertip or earlobe. US 20180317787 A1 describes a reflection-type pulse sensor for heart rate monitoring, which emits infrared, red, or green light (~550 nm) towards the body and measure the amount of reflected light using a photodiode or phototransistor. Oxygenated haemoglobin present in arterial blood has the property of absorbing incident light, so by measuring the blood flow rate (change in blood vessel volume), which changes overtime as a result of heart contractions, it is possible to measure the pulse wave signal. In other words, the amount of light absorbed will vary based on changes in blood vessel volume, resulting in a certain waveform. In addition, since reflected light is measured, the range of suitable areas is not limited, as is the case with transmissiontype pulse sensors.

[0014] A pulse plethysmography (PPG) method can be eithertransmission or reflection. Transmission plethysmography measures the penetration of LED light into body tissue and is usually suitable for examining fingertips or earlobes. Reflectance plethysmography, which places an LED next to a photodiode, continuously records scattered light from tissue and is typically suitable for examining the forehead or chest. The PPG signal is the cumulative result of various effects caused by the activity of different organ systems. The intensity of light collected by a photodetector is influenced by many factors, such as blood volume, the movement of blood vessel walls, and the orientation of red blood cells. There is a direct connection between the PPG signal and the human lungs and heart, and this connection is already being used clinically.

[0015] In general, one of the main disadvantages of photoelectric pulse wave sensors is that pulse wave measurements using red or infrared light are affected by infrared rays contained in sunlight (i.e., outdoors), which prevents stable operation. For this reason, it is recommended to use indoors or semi-indoors. For outdoor pulse wave measurement, such as with smart watches, green light source is preferred, which has high absorption rate by haemoglobin and less susceptibility to ambient light, so US 20180317787 A1 suggests using green LEDs as transmitting light sources. Other disadvantages of the PPG sensors and how the method of the present invention can overcome them are discussed in the description of the invention.

[0016] Thus, despite the usefulness of the above methods for measuring heart rate, there is an unmet demand for new types of sensors based on a different detection principle. The new sensors could, for example, be in direct contact with a part of the body, such as the limbs, and provide a rapid single-point cardiorespiratory measurement. Even more demand exists for sensors that do need to be in contact with a body but can remotely monitor the cardiorespiratory activity.

[0017] The methods described in the prior art, such as those disclosed in US 2020 / 0305740 A1 , US 2021 / 169429 A1 , and US2017 / 0112445 A1 , refer to a technique called Capacitive Plethysmography (CPG). This technique is used to measure the volume changes in biological tissues by detecting variations in capacitance based on the principle that thedielectric properties change as blood volume fluctuates. This method is very similar to PPG that measures volume change in blood using opticalsignals. For example, paragraph

[0033] of US 2017 / 0112445 A1 mentions about “arrival of blood at a fingertip causes a change in dielectric constant of the fingertip”.

[0018] US 2020 / 0305740 A1 refers to measurement of pulse wave dynamics using capacitive transducers. Fig.2A of US 2020 / 0305740 A1 refers to the placement of the electrode directly on top of an artery to detect pulse wave and hencethe change in dielectric constant. Fig 24A of US 2020 / 0305740 A1 shows the sensor response in synchronisation with PPG signal since both CPG and PPG measures the same pulse wave.

[0019] US 2021 / 169429 A1 refers to the measurement of changes in tissue capacitance during a ‘touch event’, again refers to the same Capacitive Plethysmography method. Furthermore, Fig.5B of US 2021 / 169429 A1 shows a time synchronised PPG and capacitive touch signal measurement, meaningthat it refers to a pulse wave measurement associated with the blood flow.

[0020] All the three references points to measurement of capacitance change due to the blood flow underthe electrode associatedwith heartbeat. The present invention differs from the prior art in that it describes detecting perturbations in the electric displacement field using a capacitive sensor. Human body is a conductor and can accumulate electric charge. This static charge accumulation is due to reasons such as friction while moving, touching, rubbing against surfaces etc, or it can be induction from nearby electric fields or from biological electrical activity. The amount of charge and its effects depends on the environment and interactions. This static charge accumulation is generally called as electrical potential of the human body and is discharged when connected to the ground.

[0021] The charge accumulated in the human is not a constant and it fluctuates due to induction and biological electrical activities, such as nerve impulses or muscle contractions, or biomechanical movements, such as those associated with respiration. The present invention uses a capacitive sensor to detect this charge fluctuation in the human body. When charge accumulates on human body, it affects the voltage across the connected capacitor. When the capacitor is in contact or with very proximity to the humanbody, and the static charge accumulated on the human body fluctuates due to bioelectric and / or biomechanical activities, e.g., electrical activity of the heart, muscle contractions, or breathing mechanics, a charge redistribution occurs across the capacitor plates which can be measured usingthe capacitance sensor.

[0022] Thus, the above acknowledged references measure the heart rate due to the changes in the dielectric properties of the capacitive sensor (the blood volume change due to pulse wave) whereas the present invention is measuring the fluctuations in the charge accumulated on the body due to bioelectric signals originating from, for example, the heart, or due to charge redistributions caused by biomechanical movements like respiration. Heart is a charged body in the sense that it generates and propagates electric signals due to ionic charge movements. The static charge that is accumulated on the body fluctuates when the heart, which is also an electrically charged body pumps blood out of it, or when breathing causes movement of the chest and organs. So, in the measurement of heart beats performed by the present inventors, they are measuring the electromechanical movements associated with cardiorespiratory function, such as heart beats and breathing, and this signal is superimposed on the body potential (static charge accumulated on the body).

[0023] The advantage of such a method is that it can measure the electric signals originatingfrom the heart, not just the pulse wave, which is volume change of blood. It can also detect signals related to breathing dynamics. This method also does not require the sensor to be directly placed on top of the skin and it is an ‘anywhere on the skin’ method. This method can also be used to measure electrical or mechanical signals originating from various physiological processes, including signals related to both heart function and breathing dynamics.

[0024] The present invention aims to overcome the limitations of prior art methods by providing a more accurate and versatile method for measuring cardiorespiratory activity. The invention is based on detecting fluctuations in the charge accumulated on the body due to the electromechanical activity of the cardiorespiratory system and other organs.SUMMARY

[0025] The present disclosure provides devices, systems, and non-invasive methods directed towards measuring cardiorespiratory activity and enabling associated health assessment by detecting perturbations in the electric displacement field proximate to a human or animal subject. The underlying sensing principle involves detecting fluctuations in static charge accumulated on the subject's body, which are modulated by cardiorespiratory functions. This approach distinguishes the invention from prior art techniques, such as Capacitive Plethysmography (CPG) or Photoplethysmography (PPG), which typically rely on measuring changes in tissue dielectric properties or blood volume related to pulse waves.

[0026] In one aspect, the disclosure provides a capacitive sensing device for measuring cardiorespiratory activity of a human or animal subject. The device comprises a capacitive sensor comprising means for detecting fluctuations in an electric displacement field originating from the subject's cardiorespiratory activity, wherein said fluctuations are caused by changes in static charge accumulated on the subject's body, wherein said changes are modulated by bioelectric signals or biomechanical movements associated with said cardiorespiratory activity, and means for generating measurement data representing the detected fluctuations, wherein the measurement data comprises a sequence of capacitance values, and a processor including means for processing the measurement data to determine a cardiorespiratory parameter.

[0027] In some embodiments, the capacitive sensor comprises a capacitance-to-digital converter (CDC) and means for operating with or without direct contact with the subject's body. In a further embodiment, the CDC comprises means for operating in a single-ended ground mode.

[0028] According to certain embodiments, the processor further comprises means for receiving known activity pattern data and means for training a machine-learning model which processes measurement data to determine the parameter.

[0029] Particular embodiments encompass wearable or mobile devices (e.g., sensors, rings, watches, smartphones, tablets, earphones) incorporating the sensing device,potentially integrating the sensor into a surface, and configured for continuous or triggered measurements based on events like calls or contact.

[0030] In another aspect, the disclosure provides a system for measuring and analysing cardiorespiratory activity. The system comprises at least one capacitive sensing device according to the first aspect, and a central processing unit comprising means for receiving measurement data from the device(s), determining cardiorespiratory parameters, and analysing said parameters to assess cardiorespiratory health. In embodiments comprising a plurality of sensing devices, each may comprise means for measuring activity at different locations.

[0031] In some embodiments, the central processing unit further comprises means for comparing parameters to thresholds, generating alerts based on threshold exceeding, tracking parameter changes over time, and identifying trends indicative of health changes. In certain embodiments, the at least one sensing device is integrated into a network of wearable devices.

[0032] Further embodiments relate to systems for proactive cardiorespiratory health assessment, wherein the central processing unit comprises means for analysing data using a proactive Al model (with SL means for prediction and RL means for policy optimization) to predict future health condition likelihoods and generate alerts.

[0033] In yet further detailed embodiments, such proactive systems further comprise components like a data storage component (with means for storing data), a continuous monitoring module (with means for processing real-time data), the Al-based control module (implemented by the CPU with means for autonomous determination via RL agent with learning means and SL model with prediction means), and an alert generation mechanism (with means for providing alerts).

[0034] In yet another aspect, the disclosure provides a non-invasive method for measuring cardiorespiratory activity using the device of the first aspect. The method comprises positioning the at least one device relative to a subject (potentially multiple devices at different locations), detecting the electric field fluctuations via static charge changes, generating measurement data, and processing the data via the processor todetermine a cardiorespiratory parameter. The determined parameter, according to some embodiments, is selected from heart rate, respiratory cycle parameters indicative of respiratory dynamics, or muscle activity parameters. In certain embodiments, the positioning occurs at a single sensing point located on a limb (e.g., arm, elbow, wrist, palm, finger).

[0035] The method may further comprise, in some embodiments, transmitting the data or parameter to a remote telemedicine system. Processing, in various embodiments, involves applying a machine-learning algorithm (e.g., a neural network like a fully connected, CNN, RNN, ResNet, or attention-based network) to identify patterns. Particular embodiments involve pre-processing data with a wavelet transform or utilizing neural networks comprising LSTM layers.

[0036] Further method embodiments involve proactive Al-based analysis, including analysing parameters / historical data with an Al model, predicting future condition likelihoods, and generating alerts via the processor. The Al model utilised in the method, in specific embodiments, comprises deep layers (CNN, RNN, Transformer, MLP) and / or a hybrid SL / RL architecture.

[0037] Additional embodiments detai I configuring the SL / RL components, the RL reward function (e.g., penalizing false negatives), the RL state space, training the SL component (e.g., using labelled time-series data), and training the RL component (e.g., via interaction with a simulated / real-time environment).

[0038] In yet further embodiments, the method encompasses an Al workflow comprising stages like data collection, SL model training, RL agent training (including interaction, prediction, action / reward, policy learning), and deployment for continuous monitoring and decision-making.BRIEF DESCRIPTION OF THE DRAWINGS

[0039] Disclosed embodiments will be understood and appreciated more fully from the following detailed description taken in conjunction with the appended figures.FIG. 1 schematically shows the capacitive sensing device used in the method of the present invention.FIG. 2 shows an exemplary circuit of the CDC sensor used in the method of the present invention.FIG. 3A shows the breathing activity of a user measured using a capacitive sensing device.FIG. 3B shows the noise signal when the capacitive sensing device is removed from the user’s body.FIG. 4A shows a respiratory cycle spectrogram recorded using an embodiment of the disclosed sensing device (e.g., the CDC chip Tl FDC1004 in a single channel floating mode configuration) illustrating a relatively slow respiratory cycle of about 5 sec.FIG. 4B shows a respiratory cycle spectrogram recorded using an embodiment of the disclosed sensing device illustrating a relatively normal respiratory cycle of about 2.5 sec.FIG. 4C shows a respiratory cycle spectrogram recorded using an embodiment of the disclosed sensing device illustrating a relatively fast respiratory cycle of about 1 .5 sec.FIG. 5 shows the measurements of heart signals detected using an embodiment of the disclosed sensing device with the sampling rate of 400 Hz and the bias capacitance set at 15 pF.FIGs. 6A-6B schematically show the machine learning-based architecture and waveletbased signal processing, respectively, for compressing sensor-derived data and predicting cardiorespiratory parameters such as heart rate.FIGs. 7A-7B illustrate the comparison of the typical RNN vs LSTM.FIG. 8 shows the measured signal reflecting static charge accumulation when the sensor is not grounded and grounded.FIG. 9A shows a zoomed-in view of the static charge accumulation in the not grounded section from FIG. 8, highlightingthe feature being detected in the signal in the not-grounded state, which is consistent with the description explaining the detection of heart-related signals superimposed on the static charge fluctuations.FIG. 9B shows a zoomed-in view of the static charge accumulation in the grounded section from FIG. 8, demonstrating the absence of the heart-related signals.FIG. 10 shows another example of a signal reflecting static charge accumulation with a different subject, with and without grounding, again highlightingthe detection of heartbeats in the not-grounded state.FIG. 11 compares the disclosed capacitive sensor signal (labelled "Sensor Signal” for reference in the figure) with PPG and ECG signals, indicating that the sensor peaks are not time-synchronised with PPG.FIG. 12 shows the disclosed capacitive sensor (labelled " Sensor Signal”) measuring heart signals from the wrist.FIG. 13 shows the disclosed capacitive sensor (labelled " Sensor Signal”) measuring heart signals from the forehead.FIG. 14 shows the disclosed capacitive sensor (labelled " Sensor Signal”) measuring heart signals from the elbow.DETAILED DESCRIPTION

[0040] In the following description, various aspects of the present invention will be described. For purposes of explanation, specific configurations and details are set forth in order to provide a thorough understanding of the present invention. However, it will also be apparent to one skilled in the art that the present invention may be practiced without the specific details presented herein. Furthermore, well-known features may be omitted or simplified in order not to obscure the present invention.

[0041] The term “comprising”, used e.g., in the claims, should not be interpreted as being restricted to the means listed thereafter; it does not exclude other elements or steps. It needs to be interpreted as specifying the presence of the stated features, integers, steps or components as referred to, but does not preclude the presence or addition of one or more other features, integers, steps or components, or groups thereof. Thus, the scope of the expression "a device comprisingx and z" should not be limited to devices consisting only of components x and z, and the scope of the expression “a method comprising steps x and z” should not be limited to methods consisting only of steps x and z.

[0042] As used herein, the term "about" means there is a 10% tolerance of the mentioned or claimed value. As used herein, the term "and / or" includes any and allcombinations of one or more of the associated listed items. Unless otherwise defined, all terms (including technical and scientific terms) used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs. It will be further understood that terms, such as those defined in commonly used dictionaries, should be interpreted as having a meaning that is consistent with their meaning in the context of the specification and relevant art and should not be interpreted in an idealised or overly formal sense unless expressly so defined herein. Well-known functions or constructions may not be described in detail for brevity and / or clarity.

[0043] It will be understood that when an element is referred to as being "on", "attached to", "connected to", "coupled with", "contacting", etc., another element, it can be directly on, attached to, connected to, coupled with or contacting the other element or intervening elements may also be present. In contrast, when an element is referred to as being, for example, "directly on", "directly attached to", "directly connected to", "directly coupled" with or "directly contacting" another element, there are no intervening elements present. It will also be appreciated by those of skill in the art that references to a structure or feature that is disposed "adjacent" another feature may have portions that overlap or underlie the adjacent feature.Definition of Medical Terms

[0044] Throughout this description and the appended claims, certain terms are used to encompass specific concepts related to the invention. As used herein, “cardiorespiratory activity” refers to the combined physiological functions of the cardiovascular system (including the heart and blood circulation) and the respiratory system (including the lungs and breathing mechanics). Specifically, it encompasses the electrical and electromechanical phenomena associated with these functions, such as heart muscle contraction / relaxation and lung / chest movement during breathing, which generate detectable perturbations in the body's surrounding electric displacement field. The invention detects these activities via fluctuations in static charge accumulated on the subject's body.

[0045] A “cardiorespiratory parameter” is a quantifiable metric derived from processing the measurement data generated by detecting cardiorespiratory activity. Such parameters quantify specific aspects of the subject's heart function and breathing dynamics. Nonlimiting examples include heart rate, inter-beat intervals, respiratory rate, parameters characterising the respiratory cycle waveform, muscle activity indicators, and other statistical or morphological features extracted from the detected signals.

[0046] Furthermore, beyond cardiac and respiratory signals, the fluctuations in static charge can also be modulated by the bioelectric potentials generated during skeletal muscle contractions. Therefore, the 'muscle activity parameter' recited herein refers to a parameter derived from the measurement data by processing means specifically configured (e.g., through filtering or pattern recognition algorithms) to isolate and quantify signal components attributable to the electrical activity of muscles other than the heart, providing an indicator of general muscle engagement or specific muscle group activation.

[0047] Respiratory cycle parameter is indicative of respiratory dynamics. This term refers to a specific subset of cardiorespiratory parameters derived from the analysis of the breathing signal waveform, often visualised as a respiratory cycle spectrogram. These parameters quantify the dynamic characteristics (“respiratory dynamics”) of the subject’s breathing pattern, including, but not limited to, respiratory rate (cycles per minute), inhalation / exhalation timing and symmetry (e.g., slope analysis), presence and duration of pauses, signal amplitude, peak / trough characteristics, and signal shape features (e.g., FWHM, derivatives).

[0048] As used herein, “cardiorespiratory data” encompasses any raw or processed information related to the subject's cardiorespiratory activity obtained or utilized by the described invention. This includes the raw sequence of capacitance values generated by the sensor, processed signal representations (e.g., filtered signals, spectrograms), derived cardiorespiratory parameters, historical records and trends of such parameters, and relevant contextual data influencing cardiorespiratory function.

[0049] The term “cardiorespiratory health” refers to the overall functional state and condition of the subject’s integrated cardiovascular and respiratory systems. It is assessedbased on the analysis of current and historical cardiorespiratory data, including measured cardiorespiratory parameters, their trends over time, and comparison against normative ranges or individual baselines, to determine normal function or detect potential deterioration.

[0050] A “cardiorespiratory health condition” denotes a specific identifiable state related to the function or dysfunction of the cardiovascular and / or respiratory systems. This encompasses a spectrum from normal physiological states to various abnormalities, disorders, diseases (e.g., arrhythmias, heart failure, sleep apnoea, COPD), or acute events (e.g., myocardial infarction, stroke) that can be detected, monitored, or predicted by analysing cardiorespiratory data.

[0051] “Cardiorespiratory monitoring” refers to the process of utilising the described invention for ongoing tracking of a subject's cardiorespiratory activity and parameters over time. It involves continuous or periodic data acquisition and processing to observe the status and changes in the subject's cardiorespiratory function, often facilitated by wearable devices and potentially linked to remote systems.

[0052] The term “cardiorespiratory health assessment” describes the evaluation of a subject’s cardiorespiratory health. It may involve analysing cardiorespiratory data obtained through monitoring, identifying significant parameters or trends, comparing results against established thresholds or patterns, and potentially utilising predictive models, such as the described proactive Al, to estimate the risk of future cardiorespiratory health conditions.Clarification on Method Scope

[0053] Within this description, the fundamental method enabled by the invention is frequently referred to as a “method for measuring cardiorespiratory activity”. This terminology ('method for measuring cardiorespiratory activity') is used advisedly, as the core operational principles described constitute, in essence, a process of measurement. These core principles include positioning the unique capacitive sensing device relative to the subject; detecting the subtle fluctuations in the electric displacement field originating from the subject's heart function and respiratory movements (as reflected in changes inaccumulated body charge); generating corresponding quantitative measurement data (such as sequences of capacitance values); and performing initial processing to determine relevant cardiorespiratory parameters, for instance, heart rate, inter-beat intervals, or parameters characterizing breathing dynamics.

[0054] This act of capturing and quantifying the underlying physiological signals is distinct from the subsequent, often more complex, process of cardiorespiratory health assessment or the determination or prediction of a specific cardiorespiratory health condition. Such assessment or determination typically involves higher-level interpretation, analysis of historical trends, comparison of measured parameters against predetermined thresholds, or the application of predictive models, including sophisticated artificial intelligence embodiments discussed herein, using the parameters obtained via the foundational measurement method. While the various embodiments of the device and system described herein are clearly intended to facilitate and enable comprehensive health assessment, condition monitoring, and prediction as valuable applications and outcomes, the independent characterisation of the core non-invasive method focuses on the novel technique for reliably obtainingthe cardiorespiratory activity data in the first instance. This focus aligns with the overall inventive concept centred on a new way of measuring these physiological signals.Embodiments of the Invention

[0055] In one aspect of the disclosure, the present invention describes a non-invasive method for measuring cardiorespiratory activity of a human or animal subject at a single point or multiple points on the subject’s body with a sensing device or system comprising at least one capacitive sensor comprising means for performing the steps described herein, said method comprises:(i) positioning the at least one capacitive sensing device at a single sensing point on the subject's body or in proximity to said single sensing point with or without requiring direct contact, wherein if a plurality of said devices are present, each is positioned at a different location on the subject's body;(ii) detecting, by the capacitive sensing device, fluctuations in an electric displacement field around the subject originating from the subject's cardiorespiratory activity, wherein said fluctuations are caused by changes in static charge accumulated on the subject's body;(iii) generating, by the capacitive sensing device, measurement data representing the detected fluctuations, wherein the measurement data comprises a sequence of capacitance values; and(iv) processing, by a processor, the measurement data to determine a cardiorespiratory parameter.

[0056] As mentioned in the background, conventional electrophysiological methods, such as electrocardiogram (ECG), require the use of a reference electrode in addition to a sensitive probe used to detect electrophysical phenomena. Both probes are passive electrodes, usually made of metal, and the measured quantity, reflecting biological activity, is usually a linear transformation of the potential difference across the two electrodes.

[0057] The method and device / system of the present invention is based on the use of a capacitive sensing device. The prototype capacitive sensor used in the development stage of the present invention was a capacitive-to-digital converter (CDC), which supports singlepoint technology based on capacitive coupling effects. In this case, measurements at the single point do not require the use of a reference electrode. The CDC integrated circuit is specifically designed as means for measuring the capacitance of a sensor and means for converting that measurement into a digital format. The digital output is a numerical value that a microcontroller or computer can then use as means for making decisions, displaying information, or performing other actions. The terms “capacitive sensor”, “CDC sensor” and “CDC-based sensor” are used in the present invention interchangeably to refer to the sensor component comprising means for detecting the electric field perturbations described herein.

[0058] Nowadays, CDCs are essential in numerous applications, including touch sensing for touchscreens, touchpads and buttons, proximity detection in object detection and distance measurement, liquid level sensing for fluid level monitoring in tanks andcontainers, material Identification for distinguishing materials based on their dielectric properties, humidity sensing for measuring air moisture content, and in medical devices for medical sensing and instrumentation.

[0059] The technology behind interactive touch screens is continually evolving. Capacitive is the leading type of touchscreen technology. From smart phones and tablets to video game consoles and smart appliances, countlesstouchscreen devices are powered by capacitive technology. Along with touch points, software usage, etc., there has been significant advancement from infrared touch to the use of capacitive touch technology. In capacitive touch screens, a uniform electrostatic charge is applied to the top layer. Because the human body conducts electricity, touching the display absorbs some of that current, which is used by devices to detect touch commands. The sensitivity and accuracy of this technology are great. The glass of the screen is very durable and should remain in good condition as it registers touches.

[0060] As mentioned above, CDCs operate by applying an excitation signal to one plate of a virtual capacitor while simultaneously measuring the charge stored therein. They also make the digital result available to an external host. CDCs can differentiate between four types of capacitive sensors by changing the way the excitation is applied. By changing the values of these parameters and / or observing changes in their values, CDC technology directly measures capacitance values. The distance between the two electrodes affects the output power of the CDC in inverse proportion.

[0061] In recent years, the healthcare industry has seen many advances, innovations, and improvements in electronic technology. Medical devices have faced challenges such as developing new treatments and diagnostics, providing home health care, remote monitoring, increasing flexibility, improving quality and reliability, and increasing ease of use.

[0062] The broad portfolio of these technologies includes digital signal processing, mixed signal and linear technologies, MEMS, which have helped change the medical device landscape in areas such as patient monitoring and imaging. Another example is capacitance-to-digitiser technology providing highly sensitive capacitance measurement inhealthcare applications. For example, a capacitive touch sensor is a new user input method that can take the form of a slider, button, scroll wheel, or other similar shapes.

[0063] In a typical touch sensor layout, the circuit board may have a geometric region representing the sensor electrode. This area forms one plate of the virtual capacitor, and the user's finger forms the other plate. For this system to work, the user must be grounded relative to the sensor electrode or the system must account for floating potential, as discussed herein.

[0064] Analog Devices has developed a family of capacitive touch controller chips, the to activate and interact with capacitive touch sensors. Controller chips measure changes in the capacitance of single-electrode sensors, generating control signals to charge the capacitor plates. When another object, such as a user's finger, comes close to the sensor, it creates a virtual capacitance, with the user acting as the second plate of the capacitor. The CDC or capacitance-to-digital converter in such chips comprises means for measuring the change in capacitance.

[0065] In addition, CDC can comprise means for measuring changes in the capacitance of external sensors and use this information to activate the sensor. Some controller chips have built-in calibration logic comprising means for compensating for measurement changes due to changes in ambient temperature and humidity, thereby ensuring that there are no false positives due to such changes. Such CDCs offer multiple operating modes, highly flexible control functions and user-programmable conversion sequences. These features make CDC ideal for high-resolution touch sensors that act like scroll wheels or sliders and require minimal software support. Likewise, applications with touch buttons and embedded digital logic do not require any software support.

[0066] CDCs use various techniques to achieve capacitance-to-digital conversion. One of the most common methods for operating CDCs is a voltage-based charge transfer method, where the sensor capacitor is charged to a fixed, known voltage, and the charge on the sensor capacitor is transferred to a reference capacitor with a known capacitance. The resulting voltage on the reference capacitor is then measured. This voltage is directly proportional to the ratio of the sensor's capacitance to the reference capacitance. Ananalogue-to-digital converter (ADC) then comprises means for digitising this voltage, providing the digital representation of the capacitance.

[0067] Another CDC method is an oscillator method utilising an oscillator circuit, wherein the sensor capacitor is integrated into a relaxation oscillator circuit (where the capacitor charges and discharges repeatedly). The oscillation frequency of this circuit is directly affected by the sensor's capacitance. An electronic counter comprises means for measuring either the number of oscillations in a fixed time period or the time for a fixed number of oscillations. The frequency or count is then used as means calculate the capacitance, providingthe digital output.

[0068] A time-based measurement method of operating CDCs is based on charging the sensor capacitor with a known constant current. The time it takes for the capacitor to reach a specific voltage threshold is measured. This charging time is directly proportional to the calculated capacitance. The charging time can then be translated into a digital form representing the capacitance.

[0069] The current approach to heart rate measurement in most consumer wearable devices involves the use of pulse plethysmogram (PPG), an optical measurement method. However, this method has serious limitations when applied to wearable devices. Given its optical nature, to accurately measure heart rate the device must be in direct contact with the skin, usually on the wrist or finger. Additionally, PPG technology tends to drain the battery significantly, primarily due to the use of multiple high-intensity LEDs.

[0070] Moreover, movement can cause significant noise in the PPG signal, making it difficult to accurately extract heart rate or other data, which is clearly a major limitation for applications like exercise monitoring. Also, the PPG technology has a limited accuracy for complex measurements such as detailed respiratory analysis. While PPG is capable of measuring heart rate, extracting other parameters like blood pressure or blood oxygen saturation becomes more challenging and less reliable compared to dedicated methods like pulse oximetry or blood pressure cuffs. Similarly, extracting detailed respiratory dynamics from PPG can be very difficult.

[0071] Furthermore, skin pigmentation and blood flow variations can drastically affect the PPG signal, leading to inaccurate readings, especially in people with darker skin tones or poor peripheral circulation. Being optical in nature and therefore sensitive to scattering and reflection of incident light, PPG clearly faces challenges when applied to people with darker skin tones, tattoos or hair in the measurement area.

[0072] Last but not least, PPG sensors are very sensitive to their placement. The quality of the signal may depend on the location of the sensor on the body. Finding an optimal location with good blood flow and minimal motion artifact can be critical for accurate measurements. Green and infrared light are used in the PPG sensor because they're absorbed differently by oxygenated and deoxygenated haemoglobin. However, other chromophores (light-absorbing molecules) in the skin can also absorb these wavelengths, introducing noise into the signal, which significantly affects sensitivity of the device. Since the amount of light reachingthe detector can vary depending on sensor placement and how tightly the sensor is secured. A weak signal can make it much harder to distinguish physiological variations from background noise. Sophisticated algorithms are needed to extract meaningful data from the raw absorption measurements in the PPG-based devices. The quality of these algorithms significantly impacts the accuracy of the final results.

[0073] In view of the above, it is clear that the absorption measurements from PPG can be reliable only to a very certain extent. Although PPG offers a convenient and non-invasive way to monitor heart rate, its severe limitations, relatively low sensitivity, and how factors such as movement and skin characteristics can affect the reliability of the measurements may outweigh its usefulness, particularly for comprehensive cardiorespiratory assessment.

[0074] By addressing the aforementioned shortcomings of the PPG technology, the measurement method and system of the present invention, based on the use of CDC sensors detecting electric field perturbations, overcome the limitations associated with this and other traditional methods while providing the potentially richer cardiorespiratory signals, including heart rate and breathing dynamics. The method and system of the present invention provide therefore a non-optical direct cardiorespiratory measurement that uses electrical signals instead of optical signals to measure cardiac activity and respiratorydynamics. It should be noted that animal fur is not a concern in the method of the present invention, since it is not an optical method, in contrast to optical methods, where the fur needs to be removed and skin is exposed for optical-based heart rate measurements and monitoring devices for animals.

[0075] The key advantages of the CDCs over other sensing devices, specifically those mentioned above, are potential for high-accuracy and resolution in detecting the relevant field perturbations. CDCs comprise means for providing precise capacitance measurements with very fine resolution, detecting even the smallest changes in capacitance relevant to the underlying physiological signal. They directly output digital information, making integration into modern systems easy. In addition, CDC designs often incorporate noise reduction techniques for reliable operation.

[0076] Oscillating charges and accelerating charges create electromagnetic fields that penetrate free space and other media over long distances. This forms the basis of the proposed alternative sensing modality. By using transitions in the electronic states of the sensor modulated by local bias current perturbations resultingfrom electrophysiological or biomechanical activity, non-local readout can be achieved, which is not possible using traditional methods measuring, e.g., potential difference due to the scalar nature of the quantity measured in the latter case. The heart is an electromechanical organ from which oscillations of the electric field constantly emanate. Similarly, breathing involves mechanical movements that can perturb the local electric field. The sensor used in the method and system of the present invention is therefore designed as means for detecting local disturbances in the electric field that occur due to the constant beating of the heart and the rhythmic movements of respiration.

[0077] Thus, unlike traditional approaches that measure potential difference, a scalar quantity, the CDC sensor used in the present invention comprises means for responding primarily to disturbances in the electrical displacement field in its vicinity, which is a vector quantity. Thus, this sensor comprises means for detecting temporary fluctuations in the displacement field from a spatially fixed oscillating charge source (a type of current) inaddition to charges moving in space near the sensor (another type of current). Taken together, these two cases can be grouped as displacement currents.

[0078] Moreover, plethysmography cannot detect direct electrophysiological functions like the heart's electrical activity and must be placed directly over blood vessels to be able to detect the pulse and changes in blood volume and in the microvascular bed of the tissue. The method and system of the present invention allows the use of CDC-based sensors comprising means for detecting and determining electrophysiological and biomechanical signals related to cardiorespiratory functions in any part of the body without placing the sensor chips directly on top of blood vessels. Another problem solved by the present invention is the potential analysis of big data relating to the determination of electrophysiological and biomechanical parameters of any part of the body, in particular cardiorespiratory parameters like heart rate and respiratory dynamics. The CDC-based medical sensors operating on the principle described herein have notyet been used for big data analysis in this manner due to the aforementioned limitation of prior art CDC technology applications (e.g., CPG focusing on dielectric changes).

[0079] Reference is made to FIG. 1 schematically showing the CDC-based sensing device used in the method and system of the present invention, where (1) is a sensing electrode connected to the CDC chip (2) with an excitation source (3). A dielectric layer (4) is added on top of the sensing electrode (1). Thus, the sensing device may comprise a very sensitive capacitor, one side of which is connected to the human or animal body and the other to an analogue-to-digital converter. In its simplest form it can be embodied using a very sensitive capacitor-to-digital converter operating in a single-ended mode. To increase sensitivity, an external capacitor can be optionally added in parallel to the stabilisation circuit.

[0080] Thus, the above device shown in FIG. 1 can be configured as a CDC in single- ended ground mode. That is, the sensing device used in the method and system of the present invention can function as a grounded capacitive sensor. It may utilise a parallel plate capacitor configuration, where the sensing electrode (1) is one plate and the human or animal body acts as the other. In other words, in this setup, the sensor comprises a metalor conductor serving as one electrode of a parallel plate capacitor. The human or animal body serves as the other electrode, with the space between them constituting the dielectric layer (4). This dielectric may consist of a thin polymer layer atop the metal plate, fabric, or air. Therefore, direct physical contact is not mandatory; close proximity suffices, contingent upon the sensor's sensitivity and configuration.

[0081] To sum-up, instead of using light, the CDC-based sensor disclosed herein comprises electrodes placed close to the skin or operable non-contactly. These electrodes are part of means for creating or interacting with a very small electric field. As the heart dipole changes with each heartbeat, and as breathing movements occur, the electric field at the skin perturbates due to the resulting fluctuations in accumulated static charge. This perturbation leads to minute changes in capacitance detectable by the CDC chip comprising means for high-resolution measurement.

[0082] As mentioned above, CDCs can be incredibly power-efficient, potentially leading to longer battery life in wearable devices. There is no light interference and therefore, capacitive sensing accordingto the present principle would not be affected by ambient light conditions, which can sometimes be a problem for optical sensors. Capacitive sensors based on this principle could potentially be integrated into more flexible designs than optical PPG setups and have different form factors. Moreover, CDCs can comprise means for detecting the tiny capacitance changes induced by heart dipole variations and breathing mechanics. As a result, although capacitive sensing can be prone to noise from the environment and the user’s body, improved shielding and signal processing can be used to solve this problem.

[0083] Thus, one of the main advantages of the described CDC-based sensing device is an ability to perform very accurate single-point measurements on the subject’s body related to cardiorespiratory activity. Such sensor can be positioned anywhere on any skin or even in a close proximity to the skin. No optics is involved in the measurement, and the CDC circuit is of a very small size and very low power, which can be essentially miniaturised. Any change in the relevant aspects of the generated electric field perturbations will cause a change in the stored energy or charge distribution in the capacitor system and is immediately detectedby the CDC circuit comprising means for such detection. FIG. 2 shows an example of such circuit. The configuration can be compared to the operation of a capacitive touch sensor. Here, instead of the proximity of a finger (or human interaction) with a sensor, changes in the electric field emanating from or perturbed by the body are recorded.

[0084] This approach facilitates highly precise contactless sensing with minimal power consumption. Charges in motion, such as oscillating or accelerating charges, generate electromagnetic fields that extend into free space and various mediums across significant distances. Any electrophysiological or biomechanical activity occurring within the human body induces changes in the local displacement field and associated static charge distribution. A highly sensitive capacitive sensor, as described herein, comprises means for detecting these alterations in the local electric field environment. For instance, the heart, being an electromechanical organ, continuously produces fluctuations in its electric field. Similarly, respiration produces mechanical movements that perturb the field.

[0085] Since the human body is a charged conductor, the present invention provides means for measuring the electromechanical activity of the heart occurring within the charged body, as well as the effects of respiration on the charge distribution. Likewise, during the breath cycle (inhalation and exhalation), the chest, diaphragm and lungs move within the charged body, perturbing the surrounding electric field. The present invention provides means for measuring electrical field perturbations around the human body resulting from electromechanical activities such as contractions of the heart and mechanical movements of the lungs and chest associated with breathing. Since changes in the overall electric field around the body (as reflected in static charge fluctuations) are measured, the present invention allows measurements to be taken anywhere on the skin or even in close proximity to the skin.

[0086] Thus, the CDC sensor comprises means for measuring any capacitance between the input channel and the ground or relevant reference potential. Any dynamic parasitic capacitance can also be eliminated using shield sensor. The shield sensor is excited using the same excitation as the input channel and is designed as means for picking up fringing fields arising outside the main contact point. The shield sensor picks up the electric fieldfluctuation arising out of movement whereas the input channel will pick up both the movement and the signal of interest (e.g., cardiorespiratory activity). Signal processing means can then be used to isolate the signal of interest.

[0087] In the CDC configuration of the present invention, the capacitive sensor must possess a scale range of no less than + / -15 pF (picofarad) in order to effectively detect electrophysiological and biomechanical activity signals via the described mechanism. That is, the input range of the CDC-based sensor used in the described embodiments may be + / - 15 pF with the resolution of 0.5 pF (picofarad), although other ranges and resolutions may be suitable depending on the specific implementation.

[0088] Given that the human body can easily harbour a substantial static electric field, it is imperative for some embodiments to incorporate offset capacitance compensation of up to 100 pF. This offset capacitance needs to be dynamically regulated to account for environmental fluctuations, including static charge accumulation in the body. The offset capacitance can be managed through an external or internal capacitor, which can be modulated using a capacitance digital-to-analogue converter (CAPDAC) comprising means for such modulation.

[0089] In conclusion, the electrical activity of the heart is one of the most noticeable electrical signals generated in the body, and it is one of the simplest signals that can be measured by the method and system of the present invention. The movements of the lungs and chest during breathing cause perturbations in the electrical field environment around the human body. Thus, the change in electric field due to lung activity can be measured using the described CDC-based sensor comprising means for detecting said changes. The advantage here is that it allows for monitoring dynamic breathing.

[0090] Smartphone screens contain large amounts of CDCs. This actually allows the smartphone screen itself to be enabled or used as part of a sensor system comprising means for detecting cardiorespiratory activity, provided the CDCs are sensitive enough and appropriately configured. Since direct contact with the metal electrode is not necessary in some embodiments, it should be possible to measure activity even when the phone is in the user's pocket, depending on sensitivity and configuration.

[0091] Physically, the human heart represents a volume source of an electric dipole field acting within a volume electrolytic conductor represented by human body. Respiration involves mechanical volume and position changes. Using the enormously high charge sensitivity potentially available with CDC technology configured according to the present invention, it is possible using disclosed means to record signals related to both processes: the dynamic distribution of an electrical heart dipole due to a heart muscle polarisation / depolarisation cycle, and signals correlated with corresponding mechanical movements of the heart and respiratory system in near realtime.

[0092] In a particular embodiment, the sensor of the present invention is used for cardiorespiratory monitoring. A processor comprising means for processing and analysis, optionally part of a graphical user interface (GUI) is programmed as means for interpreting the signal peaks obtained from the sensor, means for processing their shape and time intervals, and means for correlatingthem with the corresponding peak / point readings, and with related intervals between said points in the spectra to determine cardiorespiratory parameters.

[0093] The 'processor comprising means for processing and analysis' recited herein structurally corresponds to computing hardware components such as, but not limited to, microcontrollers, Central Processing Units (CPUs), Field-Programmable Gate Arrays (FPGAs), Application-Specific Integrated Circuits (ASICs), or combinations thereof, configured by executing specific software instructions or hardware logic designed to perform the claimed processing steps, including receiving measurement data, applying algorithms for noise filtering, feature extraction, parameter calculation (e.g., heart rate, respiratory dynamics), and implementing machine learning or Al models as further described.

[0094] The sensor comprises means for measuring any capacitance between the input channel and the ground or reference. Any dynamic parasitic capacitance can also be eliminated using shield sensor. The shield sensor is excited usingthe same excitation as the input channel and is designed as means for picking up f ringingfields arising outside the main contact point. For example, the shield sensor picks up the electric field fluctuation arisingout of movement whereas the input channel will pick up both the movement and the signal. Signal processing means can then isolate the desired cardiorespiratory signal.

[0095] Reference is now made to FIGs. 4A-4C showing the respiratory cycle spectrograms recorded using an embodiment like the CDC chip Tl FDC1004 in a single channel floating mode configuration. Offset capacitance is set to 0. FIG. 4A shows the slow respiratory cycle illustrating respiratory dynamics with the cycle period of approximately 5 second. A relatively small slope on the left (up) side of the peaks indicates slow inhalation, and a relatively large slope on the right (down) side of the peaks indicates fast exhalation. FIG. 4B shows the normal respiratory cycle illustrating respiratory dynamics with the cycle period of approximately 2.5 second. Equal slope on both sides of the peaks indicates the symmetry of inhalation and exhalation. FIG. 4C shows the fast respiratory cycle illustrating respiratory dynamics with the cycle period of approximately 1 .5 second.

[0096] An effective way to implement the method and system of the present invention would be to incorporate a suitable CDC into the capacitive touchscreen of a smart device. The capacitive touch screen consists of several proximity sensors. Although they are different from the specific capacitive-to-digital converter sensor described above configured for cardiorespiratory sensing, most of the components remain the same. Therefore, it would be possible to implement the CDC functionality as part of the system so that the screen itself can act as part of the sensor means, reducingthe circuit area.

[0097] For the data analysis, the main parameters of the respiration signal that can be determined by processing means include the number of cycles per minute, the distance between peaks and troughs, the amplitude of the signal after offset correction, the slope of the inhalation and exhalation signal, and the shape of the signal (for example, a short pause during inhalation or exhalation), pause duration at the top and bottom of the peaks, full width at half maximum (FWHM), first- and second-time derivatives. These parameters help identify respiratory characteristics and may be used by analysis means identify respiratory disorders or diseases that can be detected by analysing respiratory activity. Non-limiting examples of conditions potentially detectable via analysis of such parameters are sleeping apnoea, asthma, chronic obstructive pulmonary disease (COPD), pulmonary oedema, lungcancer, etc. It is also possible to detect the respiratory activity of infants and potentially provide data useful for monitoring for sudden infant death syndrome (SIDS). The system may employ processing means, potentially including trained machine-learning methods using signals that are known to come from subjects suffering from any respiratory disease, and the trained model can be used as means for identifying respiratory disorders.

[0098] FIG. 5 shows the measurements of heart signals detected usingthe same system but with different parameters, potentially showing superimposed heart and breathing signals. Here the sampling rate is 400 Hz and the bias capacitance is set at 15 pF. In this configuration, the system comprises means enabling detection of overlapping heart and lung signals.

[0099] In a further embodiment, the CDC-based sensors used in the method of the present invention can be applied to a single-sensing point on subject's limbs (arms, elbows, forearms, wrists, palms or fingers) or can be used remotely from a subject's body. In a particular embodiment, the sensor and system of the present invention can be used for cardiorespiratory monitoring and detection of cardiorespiratory parameters including heart rate from any single point on a subject's body and specifically from the wrist. The sensor comprises means for application to detection of electrical field perturbations or changes in electrical field near a body skin as a result of cardiac electromechanical dynamics (e.g., heart dipole charge dynamics during a complete heart cycle from atrial depolarisation to ventricular repolarisation) and respiratory mechanics. In yet further embodiment, the sensor of the present invention comprises means allowing contactless operation and use for remote cardiorespiratory monitoring and measuring cardiorespiratory parameters including heart rate.

[0100] In a certain embodiment, the measurement data is processed in the processor of a sensing device by applying a machine-learning (ML) method. The measurement data comprises a string or array of classical bits representing capacitance measurements with time from each capacitive sensor in the sensing device. In other words, the input to the ML method may be derived from a stringer array of classical bits (in case of appearance a single signal or peak in the spectra at a certain time event point, which is time of heart beat) orintegers (in case of a group of signals) yifwhere each bit in the string accepts the value of either 'O' or '1 ' corresponding to a non-successful and successful measurement of the signal amplitude corresponding to capacitance measurement at time(time stamp), respectively, and where the value of each integer is any integer number correspondingto the measured frequency, voltage, electrical resistance, capacitance, maximum availability or amplitude of the input signal features. This is particularly relevant when processing data representing multiple cardiorespiratory cycles, many capacitance measurements, which are performed via a series of experiments, and therefore, the generated heart rate data is an average of many peaks in the capacitance spectra in a form of a continuous output.

[0101] The ML algorithm comprises means for searching for peaks or other relevant features in this input and means for tagging each peak or feature to a physiologicalevent like a heartbeat or breath cycle. Adaptive thresholds may be used by processing means to distinguish true peaks or features from noise in the signal. The algorithm may then comprise means for calculating the time between successive peaks (e.g., inter-beat intervals or breath-to-breath intervals), and the heart rate will be calculated e.g., as beats or breaths per minute, i.e., 60 to divide by average inter-beat or inter-breath intervals in seconds.

[0102] All complex data set obtained from the sensor measurements can be used as a sequence input to machine learning means for generating the predicted heart rates and / or other cardiorespiratory parameters of various human subjects or users. Details of inputgeneration, hidden-layer optimisation and data-compressing using wavelet-based signal processing are provided below. FIGs. 6A-6B schematically show an exemplary machine learning-based architecture and wavelet-based signal processing, respectively, which may be used as means for compressing the CDC-derived large data set representing the results obtained from different subjects and different single point measurements for predicting the heart rate or other parameters.

[0103] Wherein the processor comprises 'means fortraining a machine-learning model' or wherein processing involves applying a 'machine-learning algorithm implemented by a neural network', these means structurally correspond to the processor executing specific machine learning algorithms and neural networkarchitectures (such as the fully connected,CNN, RNN, ResNet, LSTM, or attention-based networks explicitly discussed) which are trained using known cardiorespiratory pattern data. The 'means for processing the measurement data' within the machine-learning model corresponds to the trained model, residing in memory and executed by the processor, applying its learned parameters to new measurement data to determine the cardiorespiratory parameter.

[0104] The ML neural network architecture shown is implemented using a MATLAB program with a typical sequence input layer fed from time-resolved data. This particular time-capacitance sequence is further processed at the LSTM and fully connected layers. Then, finally, at the end of the process, a regression layer is added as means for outputting the predicted heart rate (or other parameter) for a specific subject at a specific single point on the subject’s body. Due to the large data set, wavelet-based signal processing may be used as means for compressing the time sequence data (N) from the approximated coefficient values (N / 2), which are then used for the sequence input layer of the deep neural network.

[0105] The abbreviation “LSTM layers” refers to the long short-term memory layers of a neural network. Recurrent neural networks (RNNs) process input data in a sequential manner, with the context of the previous input taken into account when calculating the output of the current step. This allows the neural network to transmit information over different time steps rather than keeping all inputs independent of each other. However, a significant drawback that RNNs face is the problem of vanishing / exploding gradients. This problem occurs when backpropagating a signal through a typical RNN during training, especially for networks with deeper layers. Gradients must undergo continuous matrix multiplication during the backpropagation process due to the chain rule, causing the gradient to either shrink exponentially (disappear) or expand exponentially (explode). Gradients that are too small prevent the weights from being updated and trained, and gradients that are extremely large cause model instability. Due to these problems, RNNs are unable to process longer sequences and retain long-term dependencies, causing them to suffer from “short-term memory”. LSTMs are designed as means for solving this problem.

[0106] Although LSTMs are a type of RNN and function similarly to traditional RNNs, they differ in the gate mechanism. This feature solves the "short-term memory" problem of RNNs. FIGs. 7A-7B illustrate the comparison of the typical RNN vs LSTM. As clearly seen from these figures, the difference lies mainly in the ability of LSTMs comprising means for retaining longterm memory. This is especially important in most sequential tasks that use timestamps, such as processing cardiorespiratory data. The neural network processing means of the present invention can generate estimations of heart beats or maximum availability values based on the capacitance input given to the network. At the start of the input, the network receives information about certain peaks or features in the capacitance spectra when certain heart beats or breaths appear at certain time stamps. Due to short-term memory, a typical RNN will only be able to associate these new peaks or features with recently detected events, which is completely useless for long-range dependencies, and it has no clue as to what the single point measurements relating to other peaks or features might be because relevant information from the beginning of the input has already been lost.

[0107] On the other hand, LSTM layers comprise means for storing earlier information about all previous timestamps, and this will help the model accurately identify heart rate and / or respiratory dynamics in complex situations, even if the subjects or users are physiologically very similar or single points of measurements are very close. LSTM is capable of generating input data at any timestamp thanks to contextual information from a much earlier timestamp. This essential advantage of LSTM actually lies in its gating mechanism within each LSTM cell. In a typical RNN cell, the input at a certain time stamp and the hidden state from the previous time step are passed through a hyperbolic tangent (tanh) activation function to produce a new hidden state and output.

[0108] On the other hand, an LSTM cell at each timestamp receives three different pieces of information: the current input data, short-term memory from the previous cell (similar to hidden states in an RNN), and finally long-term memory. Short-term memory is usually called the hidden state, and long-term memory is usually called the cell state. The cell then uses gates to regulate the information that should be retained or discarded at each time step before passing long-term and short-term information to the next cell.

[0109] These gates can be thought of as filters. Ideally, the role of these gates is to selectively remove any unnecessary information. At the same time, only relevant data can pass through these filters, just as gates store only useful information. Of course, these gates need to be trained as part of the ML model training process to accurately filterwhat is useful and what is not. These gates are called the Input Gate, the Forget Gate, and the Output Gate. There are many variations in the name of this gate; however, the calculations and operation of these gates are basically the same.

[0110] An output of a neural network processing means used in the method and system of the present invention may be defined as a single bit whose value is 'O' or '1 ', or an array of bits, or an array of integers, or an array of complex numbers, wherein said single bit, or said array of bits, or said array of integers, or said array of complex numbers corresponds to an estimated frequency, voltage, electrical resistance, capacitance, inter-event time intervals (e.g., inter-beat or inter-breath intervals), and to maximum availability or amplitude of the input signal features. As mentioned above, a neural network means of the present invention is trained using a training dataset of inputs with known cardiorespiratory patterns. During the training, the parameters of the neural network are optimised using known methods to output the correct cardiorespiratory patterns or parameters for the inputs in the training dataset. The goal of the training is thus to make the neural network learn the general relation between the inputs and outputs such that it would be able to output the correct result for a previously unseen input with the highest possible probability.

[0111] The machine-learning method of the present invention is able to overcome the lack of knowledge of the physical model under supervised learning. As noted above, the objective is to use a train data set that contains inputs together with their known true cardiorespiratory patterns (true outputs) in order to train a deep neural network such that the trained neural network is an optimised function, which outputs the correct results for new inputs with the optimal (or near optimal) probability.

[0112] Thus, the problem of discrimination between meaningful physiological signals corresponding to cardiorespiratory events and noise or artifacts by the CDC sensor used in the method and system of the present invention is solved using such processing means. Inaccordance with the present invention, to overcome the model's lack of knowledge, a supervised machine-learning model may be used. In this model, a train dataset of measurement results of known cardiorespiratory patterns from a pre-generated library may be used to train the neural network. The trained network is then applied to a test dataset and results in estimations of the relevant cardiorespiratory parameters, such as time stamps of the test measurement results to finally output the correct inter-beat intervals or heart rate calculated from them, or corresponding respiratory parameters.Proactive Al Embodiments

[0113] In a further embodiment, the present invention incorporates a proactive artificial intelligence (Al) system comprising means to enhance its capabilities beyond real-time monitoring and extend into predictive health assessment. This proactive Al model comprises means for analysing the measured cardiorespiratory parameters and historical data and means to predict the likelihood of future health conditions, enabling early detection and intervention and potentially improving patient outcomes.

[0114] By definition, proactive Al is an artificial intelligence system that anticipates and fulfils the needs of users without prompting. This type of Al uses the latest algorithms, machine learning (ML), and predictive analytics as means to forecast future commands and proactively respond to them. Reactive Al, in contrast, operates in the moment. It analyses incoming data, compares it to its knowledge base, and then reacts accordingly. It can be thought of as a reflex action. It does not predict or anticipate future events; it simply responds to the current situation.

[0115] Proactive Al, on the other hand, anticipates future needs and events. It leverages historical data, predictive models, and real-time information using means for making decisions and taking action before an event occurs. This allows it to optimise processes, prevent problems, and capitalise on opportunities. Reactive Al focuses on reacting to current situations, whereas proactive Al anticipates and prevents future events. Reactive Al uses primarily current data and makes decisions based on immediate context, whereas proactive Al uses historical, real-time, and predictive data and makes decisions based onpredictions and long-term goals. Thus, reactive Al only responds to events, whereas proactive Al initiates actions.

[0116] For example, a sensor with reactive Al merely observes a single parameter, compares it to a set threshold, and triggers an alert. However, the proactive Al embodiments of the present invention offer a significant advancement. The Al system comprises means for analysing historical sensor data and means for generating performance logs and also comprises means for incorporating external factors like climate, temperature, atmospheric pressure, time of day, and fatigue into its analysis. This holistic approach enables the system comprising means for predicting and proactively alerting the user to impending health deterioration.

[0117] The proactive Al model is preferably a hybrid model combining its two core components: reinforcement learning (RL) agent and supervised learning (SL) model. This hybrid approach leverages the strengths of both methodologies using means for achieving robust and adaptive predictive capabilities.Reinforcement Learning (RL) Agent

[0118] The RL agent's goal is to learn the optimal policy for analysing complex cardiorespiratory data patterns and generating timely and appropriate alerts or interventions. Its state encompasses a rich representation of the subject's cardiorespiratory status, including:- Real-time sensor data: High-resolution cardiorespiratory parameters derived from the capacitive sensor, such as heart rate variability (HRV), ECGwaveform features (e.g., QRS complex morphology), pulse wave velocity, and respiratory rate patterns.- Historical cardiorespiratory data: The subject's longitudinal cardiorespiratory data, including trends and deviations from baseline values, and responses to previous interventions.- Contextual data: Information relevant to cardiorespiratory health, such as activity level, sleep patterns, medication adherence, stress levels (inferred or measured), and environmental factors (e.g., temperature, air quality).- Predicted health risks: Probabilistic predictions of future cardiorespiratory events provided by the SL model.

[0119] The action space consists of a nuanced set of decisions on the type, urgency, and timing of an alert or intervention that the system comprises means for generating, encompassing:- Informational alert: Providing feedback on trends or deviations from healthy ranges.- Advisory alert: Recommending lifestyle changes or medical consultation.- Emergency alert: Triggering immediate medical assistance.- Adjusting monitoring parameters: Increasing the frequency of data acquisition or focusing on specific parameters.

[0120] The reward function is crucial for guiding the RL agent's learning process via suitable training means. It reflects the clinical significance of the agent's actions and incorporates:(i) Penalties: False negative: High penalty for failing to predict and alert before a critical event, reflecting the severe consequences of missed diagnoses. False positive: Smaller penalty for generating unnecessary alerts, balancing the need for sensitivity with the desire to minimise alarm fatigue.(ii) Rewards: Correct prediction: Positive reward, with the magnitude of the reward increasing with the lead time of the prediction (i.e., earlier accurate predictions are rewarded more). Appropriate intervention: Positive reward for triggering the correct type of alert or intervention based on the predicted risk.(iii) Other factors: Penalties for delayed alerts, reflectingthe importance of timely action. Rewards for minimising the frequency of emergency alerts while maintaining patient safety.Supervised Learning (SL) Model

[0121] The SL model's objective is to provide means for predicting the likelihood or risk of future cardiorespiratory events based on the available data. This model provides crucial probabilistic forecasts that inform the RL agent's decision-making means. Historical data from the subject is used as training data, including: time-series data of cardiorespiratory parameters, relevant contextual data, and labelled data on the occurrence and timing of cardiorespiratory events (e.g., diagnoses, hospitalisations).

[0122] The model type can be selected based on the specific prediction task: either regression model (e.g., neural network, support vector regression) designed as means for predicting continuous risk scores or probabilities of events within a given time window, or classification model (e.g., neural network, random forest) designed as means for predicting discrete categories of risk (e.g., low, medium, high) or the likelihood of specific event types.

[0123] The various 'means' recited in the proactive Al embodiments, such as 'means for analysing', 'means for predicting likelihood', 'means for optimising a policy', 'means for learning an optimal policy', 'means for autonomously determining likelihood', and 'means for providing the alert', structurally correspond to the central processing unit (or dedicated Al co-processors like GPUs or NPUs) executing the specific Al models, components, and workflows described herein. This includes executing the software implementations of the SL model (e.g., regression or classification models based on architectures like CNNs, RNNs, transformers, MLPs) trained on historical and real-time data, the RL agent (e.g., implemented as a Deep Q-Networkorsimilar) interactingwith its environment and updating its policy based on the defined reward function, and the algorithms governing the overall Al workflow including data collection, SL training, RL training, and deployment stages. The 'data storage component' corresponds to physical memory (e.g., RAM, flash memory, disk drives), and the 'alert generation mechanism' corresponds to hardware (e.g., display, speaker, vibration motor) and software interfaces controlled by the Al module to output alerts.

[0124] The Al workflow for proactive cardiorespiratory health assessment may comprise the following stages implemented by system or method means:Step 1 (Data collection): Gathering extensive and high-quality data from the capacitive sensors and other sources, comprising: real-time cardiorespiratory parameters from continuous or frequent measurements of a comprehensive set of parameters; historical cardiorespiratory data from longitudinal records of the subject's cardiovascular parameters, ideally spanning months or years; relevant contextual data from detailed information on factors influencing cardiorespiratory health; and labelled data on cardiorespiratory events from accurate and complete records of diagnoses, hospitalisations, interventions, and other relevant events, with precise timestamps.Step 2 (Supervised learning model training): Training the SL model using the collected data via appropriate training means to accurately predict the risk or likelihood of future cardiorespiratory events. Model selection, hyperparameter tuning, and validation techniques are employed to optimise performance.Step 3 (Reinforcement learning agent training): The RL agent interacts with a simulated or real-time cardiorespiratory monitoring environment via training means. It receives the current cardiorespiratory state as input, including sensor data and risk predictions from the SL model. It takes actions (generates alerts or refrains from alerting) based on its current policy. It receives rewards and penalties based on the consequences of its actions, guided by the reward function. The RLagent learns the optimal alert policy through iterative trial and error, refining its decision-making process via learning means to maximise long-term rewards.Step 4 (Deployment): The trained RL agent is deployed in the cardiorespiratory monitoring system as means for providing proactive health assessment. The system continuously monitors the subject's cardiorespiratory parameters, utilises means (the SL model) for predicting risks, makes real-time decisions on alert generation and intervention strategies, and comprises means for adapting its policy based on new incoming data and feedback.

[0125] The hybrid Al model offers several key advantages for cardiorespiratory health monitoring:❖ Adaptability: The RL agent's ability to learn and adapt enables the system comprising means for personalisation personalise its predictions and interventions to individual variations in physiology, lifestyle, and changing health conditions, improving accuracy and relevance.❖ Proactive Prediction: The SL model’s predictive capabilities allow the system, which comprises means for prediction, to anticipate future cardiovascular events, providing valuable lead time for proactive interventions and potentially preventing adverse outcomes.❖ Optimised Alerting: The RL agent's decision-making policy provides means for optimising the timing, frequency, and severity of alerts, balancing the need for sensitivity to detect critical events with the importance of minimising alarm fatigue and unnecessary interventions.❖ Personalisation: The model can be personalised to an individual's risk profile, historical data, and specific health goals, tailoring the monitoring and intervention strategies to their unique needs using personalisation means.❖ Comprehensive Assessment: The integration of diverse data sources (sensor data, historical records, contextual factors) enables a more comprehensive and holistic assessment of cardiorespiratory health via suitable analysis means.

[0126] The proactive Al model comprises means enabling training to predict the likelihood of occurrence or risk of a range of future health conditions, including but not limited to: stroke (e.g., ischemic or haemorrhagic stroke), myocardial infarction (heart attack), atrial fibrillation and other arrhythmias, heart failure exacerbation (worsening of heart failure symptoms), critical hypotensive events (dangerous drops in blood pressure), sudden cardiac death risk, and increased risk of cardiovascular or cardiorespiratory mortality.

[0127] In a further embodiment, the proactive Al comprises at least one of the following deep layers within its architecture selected from the group comprising Convolutional Neural Networks (CNNs) for waveform analysis (e.g., ECG-like or respiratory waveforms), Recurrent Neural Networks (RNNs) or Transformers for temporal pattern analysis, Multi-Layer Perceptrons (MLPs) for feature integration and contextualisation, and Reinforcement Learning (RL) Network (e.g., Deep Q-Network).

[0128] CNNs may be included as means for processing the time-series data from the capacitive sensor, specifically focusing on the portion corresponding to the electrical activity of the heart (ECG-like signal) or respiratory mechanics, to extract clinically relevant features. They comprise means enabling identification of subtle morphological changes in the waveform, such as ST-segment elevation or depression, Q wave abnormalities, or changes in QRS complex duration, or characteristic changes in respiratory waveform morphology, that may precede or indicate acute or chronic cardiorespiratory events. CNNs are adept at automatically learning and recognising complex spatial hierarchies in timeseries data and comprise means enablingtrainingto associate specific waveform patterns with various cardiac and / or respiratory conditions.

[0129] RNNs orTransformers may be included as means for handling sequential data of cardiorespiratory parameters over extended periods, capturing long-term trends and dependencies. Examples include, but not limited to heart rate variability (HRV) trends over days or weeks, blood pressure patterns and fluctuations, respiratory rate and pattern trends, changes in activity levels and sleep patterns. RNNs (especially LSTMs or GRUs) or Transformer models are well-suited for modelling time-series data, capturing temporal relationships, and identifying patterns of gradual deterioration or instability that may not be apparent in short-term analysis.

[0130] MLPs may be included as means for combining and processing features extracted by CNNs and RNNs / Transformers with other relevant information (e.g., contextual data) to create a comprehensive representation of the subject's cardiorespiratory state. MLPs may include other information, for example, on patient demographics (age, sex, etc.), medical history (diagnoses, medications), lab results (cholesterol levels, etc.), and genetic predispositions. MLPs are versatile as means for integrating and transforming data from multiple sources, learning complex non-linear relationships between various input features and the risk of future cardiorespiratory events.

[0131] RL Network (e.g., Deep Q-Network) may be included as means for receiving the integrated feature representation, contextual data, and risk predictions from the SL model as input. It comprises means for outputtingthe decision on whetherand how to generate an alert or recommend an intervention. The RL network comprises means for learning the optimal alerting policy through trial and error, guided by the reward function that carefully balances sensitivity (detecting all true events) and specificity (minimising false alarms).

[0132] In this architecture, the deep layers work together in a coordinated manner via suitable processing means as follows:- CNNs and RNNs / Transformers process the raw sensor data, extracting relevant features that capture both the instantaneous and temporal characteristics of cardiorespiratory activity.- MLPs integrate these extracted features with contextual information to create a holistic and informative representation of the subject's current and predicted cardiorespiratory state.- The RL network uses this integrated representation as means for making informed decisions about alert generation and intervention strategies.- The outcomes of the Al's actions (e.g., timeliness and accuracy of alerts, clinical outcomes, patient adherence) are fed back as rewards or penalties to the RL agent, enabling it comprising means for continuously learning and improving its policy over time (feedback loop).

[0133] To further enhance the performance and clinical utility of the proactive Al model of the present invention, CNNs and RNNs / Transformers may be pre-trained on large, diverse datasets of cardiorespiratory data (e.g., publicly available ECG databases, respiratory databases, clinical trial data). This can significantly improve their feature extraction capabilities and reduce the need for extensive training on individual patient data. Transfer learning techniques can also be employed to leverage knowledge from related domains.

[0134] Further, techniques like attention mechanisms or model-agnostic explanations (e.g., SHAP values) can be incorporated as means for providing insights into the Al'sdecision-making process, increasing transparency, interpretability, and trust among clinicians and patients. This is crucial for clinical validation and adoption.

[0135] In addition, robust safeguards and constraints are implemented as means for preventing unsafe or inappropriate alert generation, such as limiting the frequency of alerts, incorporating clinical guidelines, and allowing for clinician oversight. The Al model is designed with means to be robust to noisy or incomplete sensor data, which is common in real-world monitoring scenarios, by employing techniques like data imputation and noise reduction.

[0136] In another embodiment, the measurement of cardiorespiratory parameters including heart rate can be continuously carried out when the phone, smartwatch, smart ring, earphone or any wearable device containingthe capacitive sensors is in a contact with a hand or skin or activated on calling or when a contact is established. The relevant medical data recorded may then be transmitted via suitable communication means to a medicaldiagnostic telemedicine cloud and will be available for medical doctors. In some embodiments, the phone, smartwatch, smart ring, earphone or any wearable device containing the capacitive sensors may be used for portable long-time-operation solution within a health, fitness and remote telemedicine cloud-based diagnostics system.Distinction from Prior Art (Capacitive Plethysmography)

[0137] As mentioned in the background section of the present disclosure, the methods described in the prior art, such as those disclosed in the three acknowledged reference documents, US 2020 / 0305740 A1 , US 2021 / 169429 A1 , and US 2017 / 0112445 A1 , refer to a technique called Capacitive Plethysmography (CPG). This technique is used to measure the volume changes in biological tissues by detecting variations in capacitance based on the principle that the dielectric properties change as blood volume fluctuates. This method is very similar to PPG that measures volume change in blood using optical signals, as mentioned in paragraph

[0033] of US 2017 / 0112445 A1 saying that “arrival of blood at a fingertip causes a change in dielectric constant of the fingertip”.

[0138] US 2020 / 0305740 A1 refers to measurement of pulse wave dynamics using capacitive transducers. Fig.2A of US 2020 / 0305740 A1 refers to the placement of the electrode directly on top of an artery to detect pulse wave and hencethe change in dielectric constant. Fig 24A of US 2020 / 0305740 A1 shows the sensor response in synchronisation with PPG signal since both CPG and PPG measures the same pulse wave. US 2021 / 169429 A1 refers to the measurement of changes in tissue capacitance during a ‘touch event’, again refers to the CPG method. Also, Fig.5B of US 2021 / 169429 A1 shows a time synchronised PPG and capacitive touch signal measurement, meaning that it refers to a pulse wave measurement associated with the blood flow. All the 3 references points to measurement of capacitance change due to the blood flow under the electrode associated with heartbeat.

[0139] The present invention describes detecting perturbations in the electric displacement field using a capacitive sensor. Human body is a conductor and can accumulate electric charge. This static charge accumulation is due to reasons such as friction while moving, touching, rubbing against surfaces etc., or it can be induction from nearby electric fields or from biological electrical activity. The amount of charge and its effects depends on the environment and interactions. This static charge accumulation is generally called as electrical potential of the human body and is discharged when connected to the ground.

[0140] The charge accumulated in the human is not a constant and it fluctuates due to induction, biological electrical activities (e.g., nerve impulses or muscle contractions), and biomechanical movements (e.g., respiration). The present inventors utilise a capacitive sensor comprising means for detecting this charge fluctuation in the human body. When charge accumulates on human body, it affects the voltage across or charge distribution associated with the connected capacitor system. When the capacitor sensor is in contact orwith very proximity to the human body, and the static charge accumulated on the human body fluctuates due to bioelectric and / or biomechanical activities (e.g., electrical activity of the heart, breathing mechanics) a charge redistribution occurs in the capacitor system which can be measured usingthe capacitance sensor means.

[0141] In summary, the three published documents acknowledged above measure the heart rate due to the changes in the dielectric properties near the capacitive sensor (the blood volume change due to pulse wave), whereas the present invention provides means for measuring the fluctuations in the charge accumulated on the body due to bioelectric signals (e.g., from the heart (and / or biomechanical effects (e.g., from respiration). Heart is a charged body in the sense that it generates and propagates electric signals due to ionic charge movements. Respiration involves movement of tissues and organs. The static charge that is accumulated on the body fluctuates when the heart, which is also an electrically charged body pumps blood out of it, and / or when breathing causes movement. So, in the measurement according to embodiments of the present invention, the inventors utilise means for measuring the effects of movement of the heart (since it is a charged body) and the effects of breathing mechanics, and this signal is superimposed on the body potential (static charge accumulated on the body).

[0142] The advantage of such a method is that it can measure the electric signals originatingfrom the heart, not just the pulse wave, which is volume change of blood. It can also measure signals related to respiratory dynamics. This method also does not require the sensor to be directly placed on top of the skin and it is an ‘anywhere on the skin’ method in many embodiments. This method can also be used with appropriate sensor / system means for measuring electrical or mechanical signals originating from various physiological activities, not just the heart or lungs.

[0143] To provide evidence for the argument that the invention measures the fluctuations in the charge accumulated on the body and that the sensor is not measuring the pulse wave like in CPG, the inventors have provided additional experimental results in the Examples section.

[0144] It is crucial to understand that the detected 'changes in static charge accumulated on the subject's body' are not merely random static build-up, but rather fluctuations modulated by underlying physiological processes. Specifically, the bioelectrical activity inherent in cardiac muscle depolarization and repolarization, and the biomechanical movements associated with respiration (e.g., chest wall and diaphragmmotion), perturb the local electric field and cause subtle, dynamic redistributions of the existing static charge on the body surface relative to the sensor. The capacitive sensor, configured accordingto embodiments herein, is highly sensitive to these modulations in the electric displacement field resulting from the charge redistribution. This mechanism is fundamentally distinct from Capacitive Plethysmography (CPG), which measures changes in the dielectric properties of tissues under the sensor due to blood volume changes (pulse wave), rather than these bioelectrically and biomechanically modulated static charge effects.EXAMPLES

[0145] To demonstrate novelty and inventiveness of the present invention, and to provide robust evidence that the invention's capacitive sensor detects fluctuations in the body's accumulated charge due to bioelectric signals, rather than pulse waves like Capacitive Plethysmography (CPG), which measures blood volume changes, the inventors designed and conducted a series of targeted experiments.EXAMPLE 1The Critical Role of Grounding in Distinguishing Charge Fluctuation Measurement Experimental Setup

[0146] A highly sensitive capacitive sensor, connected to a precise measurement apparatus capable of resolving picofarad-level capacitance changes, was used. The subject was positioned comfortably to minimise movement artifacts.1) Not Grounded Condition: The subject was placed on an insulating surface (e.g., a wooden stool or a rubber mat) to electrically isolate them from the earth's ground. This allowed for the natural accumulation of static charge on their body.2) Grounded Condition: The subject was connected to the ground via a low-resistance grounding strap attached to their wrist or ankle. This provided a pathway for any accumulated static charge to dissipate to the earth, maintaining the subject at or near ground potential.Environmentalfactors (temperature, humidity, electromagnetic interference) were carefully controlled to ensure they did not influence the measurements.Experimental Procedure

[0147] The sensor was placed in close proximity to the subject's skin (without direct contactto avoid pressure artifacts). Continuous capacitance measurements were recorded for a set duration (e.g., 1 minute) in both the "not grounded" and "grounded" conditions. The recorded data was then processed to filter out high-frequency noise and enhance the signal related to cardiac activity.Results and Discussion

[0148] FIG. 8 clearly shows the overall effect of the static charge accumulation on the subject’s body when the sensor is not grounded and when it is grounded. As seen in this figure, the sensor detects a signal (heartbeat) when it is not grounded, but the signal disappears when the sensor is grounded. Thus, this figure dramatically illustrates the overall impact of grounding. In the "not grounded" state, a clear, fluctuating signal is observed. This signal exhibits a rhythmic pattern that corresponds to the subject's heartbeat. The signal's amplitude varies, indicating changes in the electric field around the sensor. In stark contrast, when the subject is "grounded," the signal almost entirely disappears, indicating the static charge drain and reducing to a baseline noise level. The sensor cannot detect fluctuations anymore The dramatic drop in signal amplitude upon grounding confirms that the sensor detects the presence of static charge on the body. This supports the argument that the sensor measures the static charge accumulated on the body.

[0149] Zoom-in-detail FIGs. 9A and 9B provide a crucial zoomed-in view of data on the static charge accumulation from FIG. 8, clearly demonstratingthe presence of a signal with heartbeats when not grounded (FIG. 9A) and the absence of a signal when grounded (FIG. 9B). It allows for a detailed examination of the signal characteristics. In the "not grounded" condition, the heartbeat-related fluctuations are clearly visible in FIG. 9A as small, repetitive oscillations, which are superimposed on a slower-varying baseline. These oscillations are consistent with the electrical activity of the heart.

[0150] In the "grounded" conditions, which are shown in FIG. 9B, these oscillations vanish, leaving only a flat line indicative of minimal electrical activity. This zoomed-in view provides compelling evidence that the sensor is indeed capturing the electrical signature of the heart and confirms that the sensor measures the fluctuations in the static charge due to the electromechanical activity of the heart. The consistent observation of a signal in the "not grounded" state and its absence in the "grounded" state confirms the robustness and reliability of the findings.

[0151] FIG. 10 replicates the experiment with a different subject, reinforcing the observation that grounding eliminates the signal. The consistent observation of a signal in the "not grounded" state and its absence in the "grounded" state confirms the robustness and reliability of the findings.

[0152] These results provide strong evidence that the sensor of the present invention is detecting the fluctuations in the static charge accumulated on the human body. When the body is grounded, this static charge is drained away, and the sensor no longer detects the heartbeat signal. This is a critical distinction from prior art methods like CPG, which measure changes in capacitance due to variations in blood volume under the sensor. Indeed, CPG signals are not expected to disappear upon grounding, as the blood volume changes are still present. This experiment provides a clear and convincing demonstration of the fundamental difference in the measurement principle of the present invention and prior art devices.EXAMPLE 2Temporal Relationship with Established Cardiac Signals (PPG and ECG)Experimental Setup

[0153] To compare the timing of the signal detected by the present invention's sensor with established measures of cardiac activity, a simultaneous recording setup was used: Capacitive Sensor: The sensor of the present invention was placed on the subject's wrist. PPG Sensor: A commercial PPG sensor, which measures blood volume changes using opticaltechniques, was placed adjacent to the capacitive sensor.ECG Sensor: An ECG sensor, which directly measures the electrical activity of the heart, was also attached to the subject (e.g., using electrodes on the chest).All sensors were synchronised to a common data acquisition system to ensure precise timing comparisons.Experimental Procedure

[0154] Simultaneous recordings were madefrom allthree sensors for a period. The data was processed to identify the timing of characteristic peaks in each signal, corresponding to heartbeats. The timing of these peaks was then compared across the three sensors.Results and Discussion

[0155] Reference is now made to FIG. 11 comparing the signal of the sensor of the invention with the signals of PPG and ECG sensors. This figure presents a multi-panel plot showing the signals from these three sensors. The ECG signal (black, bottom line) shows sharp, distinct peaks, representing the electrical depolarisation of the heart muscle. The PPG signal (top, blue line) also shows clear peaks, but these peaks are slightly delayed relative to the ECG peaks. This delay is due to the time it takes for the heart's electrical impulse to propagate to the mechanical contraction and the resulting blood flow to reach the measurement site.

[0156] Crucially, the capacitive sensor signal (red, middle line) exhibits peaks that are not time-synchronised with the PPG signal. Instead, they show a closer temporal relationship with the ECG signal, although with a less sharp morphology. This result clearly indicates that the sensor of the invention is not measuring the pulse wave (like PPG), but ratherthe electrical activity of the heart due to polarisation, which occurs slightly before the pulse wave (the appearance of the PPG peaks) and in between the R wave and T wave of the ECG signals.Thus, this experiment provides critical evidence that the present invention's sensor is not measuring the same physiological phenomenon as PPG. If it were, the peaks in the capacitive sensor signal would align with the peaks in the PPG signal.

[0157] The observed temporal offset between the measurements of the capacitive sensor and PPG signals, and the closer alignment with the ECG signals, strongly suggests that the sensor is responding to the electrical activity of the heart. This is a significant departure from the CPG sensor, which, like PPG, measures pulse wave and hemodynamic changes (in blood volume) and would therefore be expected to correlate with PPG. The fact that sensor of the invention is not time synchronised with PPG showsthat it is not measuring the pulse wave. The results therefore highlight the invention's ability to capture a different aspect of cardiac activity, providing potentially complementary information to PPG.EXAMPLE 3Spatial Independence of Measurement (Anywhere on the Skin)Experimental Setup

[0158] To demonstrate the spatial independence of the measurement, the capacitive sensor was used to record heart signals at various locations on the subject's body, representing different tissue types and distances from major blood vessels. The sensor setup remained consistent across the measurements. Locations included:- Wrist (FIG. 12): A common site for heart rate monitoring.- Forehead (FIG. 13): A bony region with less muscle mass.- Elbow (FIG. 14): Ajoint with significant muscle and tissue.Experimental Procedure

[0159] The sensor was placed gently on the skin at each location, ensuring consistent contact. A recording of the signa I was taken at each location fora set duration. The recorded signals were then analysed to identify the characteristic features of the heartbeat signal.Results and Discussion

[0160] Reference is now made to FIGs. 12-14 demonstrating that the sensor of the invention can measure heart signals from anywhere on the skin, including the wrist, forehead, and elbow. These figures show that clear heart-related signals were obtained at all three locations. While the signa I morphology may vary slightly due to differences in tissuecomposition and sensor proximity to the heart, the fundamental heartbeat pattern is consistently present. This highlights another advantage of the invention over prior art methods like PPG and CPG, which require the sensor to be placed directly over arteries.

[0161] The PPG method and traditional CPG methods rely on measuring changes associated with blood flow and therefore require placement over or in close proximity to blood vessels to obtain a strong signal. The ability of the present invention's sensorto detect heart signals at locations distant from major blood vessels demonstrates that it is not primarily measuring blood flow. Instead, it supports the surprising finding that the sensor is sensitive to the electrical activity of the heart, which propagates throughout the body. This spatial independence provides a significant advantage for practical applications, allowing for more flexible sensor placement and integration into various wearable devices.Conclusion

[0162] The comprehensive set of experiments and the detailed analysis of the results provide compelling evidence to support the following key conclusions:1) Fundamental Difference in Measurement Principle: The present invention's capacitive sensor operates on a fundamentally different principle compared to prior art methods like CPG and PPG. While prior art methods measure hemodynamic changes, specifically variations in blood volume or related dielectric properties, the present invention measures fluctuations in the body's accumulated static charge, which are modulated by the electrical activity of the heart and other bioelectrical events.2) Direct Sensitivity to Cardiac Electrical Activity: The sensor exhibits a high degree of sensitivity to the electrical activity of the heart. This is evident in its temporal correlation with ECG signals (Example 2) and its ability to detect heartbeat signals even when placed away from major blood vessels (Example 3).3) Spatial Independence and Versatility: The invention offers significant spatial independence in measurement. Effective heart signal detection is demonstrated at diverse body locations (wrist, forehead, elbow), unlike PPG and CPG, which are constrained to vascular regions. This versatility translates to greater flexibility in sensor design and integration into wearable devices.

[0163] The present invention offers a more specific measurement of the heart's electrical activity over prior art, providing considerably richer diagnostic information than indirect hemodynamic measurements. The measurement described in the present invention is also robust, that is less susceptible to variations in skin pigmentation, tissue composition, and sensor pressure, which are limitations of PPG and CPG. The spatial independence of the sensor of the present invention allows for greater flexibility in device design and integration, enabling unobtrusive and comfortable long-term monitoring. The ability to measure bioelectrical signals opens up new possibilities for monitoring other physiological processes beyond cardiac activity.

[0164] In conclusion, the present invention marks a definitive paradigm shift, charting a new course in non-invasive cardiorespiratory monitoring. It ingeniously departs from the constraints of prior art like PPG and CPG -which merely track secondary effects like blood volume or dielectric shifts - by introducing a novel and inventive capacitive sensing modality. This technology uniquely taps into the subtle perturbations in the body's own electric displacement field, driven by fluctuations in accumulated static charge that are directly modulated by the primary electromechanical and biomechanical rhythms of the heart and lungs. As compellingly demonstrated, its distinct sensitivity to underlying bioelectrical events (differing temporally from pulse waves) and its validated operational freedom - functioning effectively 'anywhere on the skin' regardless of proximity to major arteries - underscore a significant inventive leap. This robust approach not only overcomes limitations related to sensor placement, ambient conditions, and physiological variability but also promises richer, more direct physiological data. Ultimately, this invention paves the way for a new generation of versatile, accurate, and accessible cardiorespiratory health solutions, from ubiquitous wearables to sophisticated Al-driven proactive diagnostics, heralding a more insightful future for personal health management.Note on Claim Language

[0165] Regarding the language used in the appended claims, it is noted that for claims directed to the device and system (such as exemplary claims 1 -10), certain components likethe 'capacitive sensor', 'processor', 'central processing unit', 'data storage component', 'continuous monitoring module', 'Al-based control module', 'RL agent', 'SL model', and 'alert generation mechanism' are described using phrases such as 'comprising means for' or similar, followed by a recitation of the function performed by that component. Using such language is a way to draft claims in the 'means-plus-function' format, which is recognised under patent law (e.g., 35 U.S.C. § 112(f) in the US), where such format defines a component (the 'means') by the function it performs. This approach is deliberately chosen for consistency and clarity, defining these structural components by reference to the specific function or functions they are intended to perform to achieve the invention’s objectives. This distinguishes these apparatus claims from the method claims (such as exemplary claims 11 -29), which recite steps or actions. The functional definitions used in the device and system claims are intended to be understood in light of the corresponding structures, materials, or acts described herein that enable the recited functions, and their equivalents.

Claims

CLAIMS1. A capacitive sensing device for measuring cardiorespiratory activity of a human or animal subject, the device comprising:- a capacitive sensor comprising:■ means for detecting fluctuations in an electric displacement field originating from the subject's cardiorespiratory activity, wherein said fluctuations are caused by changes in static charge accumulated on the subject's body, wherein said changes are modulated by bioelectric signals or biomechanical movements associated with said cardiorespiratory activity, and■ means for generating measurement data representing the detected fluctuations, wherein the measurement data comprises a sequence of capacitance values; and- a processor comprising means for processingthe measurement data to determine a cardiorespiratory parameter.

2. The device of claim 1 , wherein the capacitive sensor comprises a capacitance-to-digital converter (CDC), and wherein the capacitive sensor comprises means for operating with or without direct contact with the subject's body.

3. The device of claim 2, wherein the CDC comprises means for operating in a single- ended ground mode.

4. The device of claims 1 or 2, wherein the processor further comprises means for:- receiving input data representing a known cardiorespiratory activity pattern; and- training a machine-learning model based on the input data, wherein the machinelearning model comprises means for processing the measurement data to determine the cardiorespiratory parameter.

5. Awearable or mobile device accordingto any one of claims 1 to 4, wherein the wearable or mobile device is selected from the group consisting of a wearable sensor, a smart ring, a smart watch, a smartphone, a tablet, and an earphone, wherein the capacitivesensor is optionally integrated into a display screen or a sensitive surface of the device, and wherein measurement of the cardiorespiratory parameter is performed continuously or is activated upon a trigger event such as initiation of a call or establishment of contact.

6. A system for measuring and analysing cardiorespiratory activity, comprising:- at least one capacitive sensing device according to claim 1 , wherein if a plurality of said devices are present, each comprises means for measuring cardiorespiratory activity at a different location on the subject's body; and- a central processing unit comprising means for:(i) receiving measurement data from the at least one capacitive sensing device;(ii) determining cardiorespiratory parameters from the measurement data; and(iii) analysing the cardiorespiratory parameters to assess the subject's cardiorespiratory health.

7. The system of claim 6, wherein the central processing unit further comprises means for:- comparing the cardiorespiratory parameters with predetermined thresholds;- generating an alert signal if a cardiorespiratory parameter exceeds a predetermined threshold;- tracking changes in the cardiorespiratory parameters over time; and- identifying trends indicative of changes in the subject's cardiorespiratory health.

8. The system of claim 6 or 7, wherein the at least one capacitive sensing device is integrated into a network of wearable devices worn by the subject.

9. The system of any one of claims 6 to 8 for proactive cardiorespiratory health assessment, wherein the central processing unit further comprises means for: a) analysing the measurement data and historical cardiorespiratory data of the subject using a proactive artificial intelligence (Al) model comprising a supervised learning (SL) component comprising means for predicting likelihood of a future cardiorespiratory health condition, and a reinforcement learning (RL) componentcomprising means for optimising a policy for generating the alert signal based on predictions from the SL component and a predefined reward function; b) predicting, using the proactive Al model, a likelihood of occurrence of a future cardiorespiratory health condition; and c) generating an alert signal based on the predicted likelihood.

10. The system of claim 9 further comprising:(1) A data storage component comprising means for storing historical cardiorespiratory data of the subject;(2) A continuous monitoring module comprising means for processing real-time cardiorespiratory data received from the at least one capacitive sensing device to extract one or more cardiorespiratory parameters;(3) A proactive artificial intelligence (Al)-based control module implemented by the central processing unit, said control module comprising means for autonomously determining likelihood of a future cardiorespiratory health condition and recommended intervention based on real-time and historical data, said control module further comprising:(i) A Reinforcement Learning (RL) agent comprising means for learning an optimal policy for generating alerts and recommending interventions considering a reward function prioritising minimising false negative alerts, while maintaining timely detection of potential cardiorespiratory health conditions; and(ii) A Supervised Learning (SL) model comprising means for predicting temporal changes or trends and the likelihood of future cardiorespiratory health conditions based on real-time and historical data, wherein the RL agent uses predictions from the SL model to make the determination and inform its decision-making process; and(4) An alert generation mechanism comprising means for providing the alert as directed by the Al-based control module.

11. A non-invasive method for measuring cardiorespiratory activity of a human or animal subject, the method comprising:- positioning the at least one capacitive sensing device of any one of claims 1 to 5 at a single sensing point on the subject's body or in proximity to said single sensing point without requiring direct contact, wherein if a plurality of said devices are present, each is positioned at a different location on the subject's body;- detecting, by the capacitive sensing device, fluctuations in an electric displacement field around the subject originating from the subject's cardiorespiratory activity, wherein said fluctuations are caused by changes in static charge accumulated on the subject's body, wherein said changes are modulated by bioelectric signals or biomechanical movements associated with said cardiorespiratory activity;- generating, by the capacitive sensing device, measurement data representing the detected fluctuations, wherein the measurement data comprises a sequence of capacitance values; and- processing, by a processor, the measurement data to determine a cardiorespiratory parameter.

12. The method of claim 11 , wherein the cardiorespiratory parameter is selected from the group consisting of heart rate, a respiratory cycle parameter indicative of respiratory dynamics, and a muscle activity parameter.

13. The method of claim 11 or 12, wherein the single sensing point is located on a limb of the subject's body.

14. The method of claim 13, wherein the limb is selected from the group consisting of an arm, an elbow, a forearm, a wrist, a palm, and a finger.

15. The method of any one of claims 11 to 14, further comprising transmitting the measurement data orthe determined cardiorespiratory parameterto a remote medicaldiagnostic telemedicine system.

16. The method of any one of claims 11 to 15, wherein the step of processing the measurement data comprises applying a machine-learning algorithm to the measurement data to identify patterns indicative of the cardiorespiratory parameter.

17. The method of claim 16, wherein the machine-learning algorithm is implemented by a neural network.

18. The method of claim 17, wherein the neural network is selected from the group consisting of a fully connected neural network, a convolutional neural network, a recurrent neural network, a ResNet neural network, and a neural networkwith attention heads.

19. The method of claim 17 or 18, further comprising processing the measurement data with a wavelet transform prior to input to the neural network.

20. The method of any one of claims 17 to 19, wherein the neural network comprises one or more Long Short-Term Memory (LSTM) layers.

21. The method of any one of claims 11 to 20, further comprising:- analysing, by the processor, the determined cardiorespiratory parameter and historical cardiorespiratory data of the subject using a proactive artificial intelligence (Al) model;- predicting, by the proactive Al model, a likelihood of occurrence of a future cardiorespiratory health condition; and- generating, by the processor, an alert signal based on the predicted likelihood.

22. The method of claim 21 , wherein the proactive Al model comprises deep neural network layers, including at least one of Convolutional Neural Network (CNN) layers, Recurrent Neural Network (RNN) layers, Transformer layers, and Multi-Layer Perceptron (MLP) layers.

23. The method of claim 21 or 22, wherein the proactive Al model comprises a hybrid architecture incorporating both supervised learning (SL) and reinforcement learning (RL) components.

24. The method of claim 23, wherein:- the SL component is configured to identify patterns in the cardiorespiratory parameter and the historical cardiorespiratory data associated with labelled instances of the future cardiorespiratory health condition; and- the RL component is configured to optimise a decision-making policy for generating the alert signal.

25. The method of claim 23 or 24, wherein the RL component utilises a reward function that assigns a higher penalty to false negative predictions of the future cardiorespiratory health condition than to false positive predictions.

26. The method of any one of claims 23 to 25, wherein the RL component operates within a defined state space, where states represent the subject's cardiorespiratory status based on the cardiorespiratory parameter and the historical cardiorespiratory data, and actions represent different alert signals.

27. The method of any one of claims 23 to 26, wherein the SL component is trained using a dataset comprising time-series data of the cardiorespiratory parameter and corresponding labels indicating the occurrence of the future cardiorespiratory health condition.

28. The method of any one of claims 23 to 27, wherein the RL component is trained through interaction with a simulated or real-time cardiorespiratory monitoring environment, receiving rewards and penalties based on the accuracy and timeliness of the generated alert signal.

9. The method of any one of claims 23 to 28, where an Al workflow of the Al-based control module comprises the following stages:A. Data collection including collecting extensive data from cardiorespiratory monitoring, including sensor readings, contextual factors, and cardiorespiratory health conditions;B. Supervised learning model training including training the supervised learning model to predict cardiorespiratory health conditions based on the collected data;C. RL agent training including:- The RL agent interactions with a simulated or real-time cardiorespiratory monitoring environment;- The RL agent receiving sensor data and using the supervised learning model to predict cardiorespiratory health condition evolution;- The RL agent taking actions of alert generation and / or intervention recommendation and receiving rewards / penalties based on prediction accuracy and timeliness; and- The RL agent learning the optimal policy through trial and error, guided by the reward function; andD. Deployment including the trained RL agent deployment on the cardiorespiratory monitoring system and continuously monitoring the cardiorespiratory parameters, making alert / intervention decisions in real-time, and adapting its policy based on new data.

Images