Methods and associated systems used to determine a subject's cardiac or respiratory activity

Abstract

The present invention relates to a method for determining the cardiac or respiratory activity of a subject, the method comprising: measuring and recording sample measurements over a period of time that are at least related to the kinematic characteristics or position of an accessory (2) worn on the head of the subject; and processing the sample measurements to determine the cardiac or respiratory activity.

Description

Technical Field

[0001] This invention relates to the field of health monitoring.

[0002] The present invention specifically relates to a method for determining the cardiac or respiratory activity of a subject, and to an associated system. Background Technology

[0003] Information on heart and respiratory activity is an important indicator of health and wellness. This information can be used for disease detection, self-health monitoring, telemedicine, stress level monitoring, driving attention level monitoring, or other applications such as exercise guidance and video games.

[0004] SCG (Seismographography) is a technique that measures chest vibrations produced by the heartbeat, using an accelerometer to record these vibrations. Compared to an electrocardiogram (ECG), SCG provides detailed information about cardiac events during each heartbeat. Figure 1 A comparison between SCG measurements and corresponding electrocardiograms is shown. It can be seen that SCG measurements provide more accurate details of cardiac activity.

[0005] Gyrocardiography (GCG) is a relatively new, non-invasive technique for acquiring and analyzing changes in chest angular velocity associated with cardiac activity.

[0006] The combination of SCG and GCG technologies allows for more accurate measurement results.

[0007] Typically, SCG / GCG measurement sensors consist of an inertial measurement unit comprised of a MEMS (microelectromechanical systems) accelerometer and a gyroscope placed on the sternum near the heart. The subject is in a supine position, or the measurement sensor is attached to the skin.

[0008] The measurements obtained can also be used to estimate the subject’s respiratory rate, respiratory volume and respiratory phase, as described in International Application WO 2020 / 037391.

[0009] The optical heart rate sensor in smartwatches uses a technique called photoplethysmography. This technique uses a green LED light and a photosensitive photodiode to detect the amount of blood flowing through the wrist at any given moment. A heartbeat corresponds to a larger absorption of green light by the blood flowing through the wrist. However, this technique is power-intensive, requires space, and is difficult to integrate mechanically. Furthermore, the sensor used for photoplethysmography is specifically designed to measure heart activity.

[0010] Therefore, there is currently no solution for low-power, small-space continuous monitoring of a subject's cardiac and respiratory activity. Summary of the Invention

[0011] In this context, the present invention proposes a method for determining the cardiac or respiratory activity of a subject, the method comprising: measuring and recording sample measurements over at least a period of time that are related to the kinematic characteristics or position of an accessory worn on the subject's head; and processing the sample measurements to determine the cardiac or respiratory activity.

[0012] Therefore, the method according to the invention aims to provide a solution for measuring the cardiac and respiratory activity of a subject in a non-constrained, permanent or semi-permanent manner.

[0013] Furthermore, compared to classic SCG / GCG measurements, this invention presents a subtle advantage: the sensor does not need to be placed in direct contact with the subject's body and close to the heart. In this invention, a tight mechanical interface is not required between the subject's body and the accessory. Therefore, this invention overcomes the bias of the prior art, which requires SCG / GCG measurements to be performed through contact between the sensor and the subject's body.

[0014] The accessory is equipped with at least one motion displacement sensor, such as an accelerometer or gyroscope, for measuring the kinematic characteristics or position of the subject's head.

[0015] Therefore, the method according to the invention uses non-invasive hardware that requires less space.

[0016] The accessories are preferably selected from the following: eyeglass frames, eyeglass attachments, eyeglass clips, eyeglass frame straps, headphones, earphones, and jewelry.

[0017] The process can be advantageously executed in an embedded manner or on a remote host.

[0018] The present invention further proposes a system for determining the cardiac or respiratory activity of a subject, the system comprising an accessory configured to be worn on the subject's head, and the accessory being provided with: a measurement unit including at least one motion displacement sensor, such as an accelerometer or gyroscope, and transmitting at least one measurement signal reporting at least the position, velocity, rotation, or acceleration of the accessory; and a processing unit programmed to perform the following steps:

[0019] -a) Receive and record at least one measurement signal from the measurement unit for at least a period of time to obtain sample measurement results.

[0020] -b) Process the sample measurements to form at least one recorded signal for use in determining cardiac or respiratory activity.

[0021] Other advantageous and non-limiting features of the system according to the invention for determining the cardiac or respiratory activity of a subject are:

[0022] - Accessories are selected from the following: eyeglass frames, eyeglass attachments, eyeglass clips, eyeglass frame straps, headphones, earphones, and jewelry;

[0023] -The processing unit is further programmed to perform the following steps before recording at least one measurement signal:

[0024] - Determine whether the sample measurement results from the measurement unit meet the triggering conditions.

[0025] - Under the condition that the triggering conditions are met, increase the sampling rate, sampling duration, and sensitivity of the measurement unit.

[0026] - Trigger the recording of at least one measurement signal,

[0027] The recording of at least one measurement signal is performed over a time slot included in the at least one period of time for obtaining the sample measurement results that form at least one recorded signal;

[0028] - Step b) includes a preprocessing stage with the following steps:

[0029] - Convert the timestamps of the sample measurement results to the selected time unit.

[0030] - The sampling frequency of the at least one recorded signal is downsampled to obtain at least one downsampled signal.

[0031] - Apply at least one window function to the at least one downsampled signal to obtain at least one windowed signal.

[0032] - Select at least one windowed signal from the at least one windowed signal.

[0033] - Estimate the quality of the at least one selected windowed signal by searching for peaks in the at least one selected windowed signal;

[0034] - When a peak is detected during the quality assessment step, the preprocessing stage further includes a step of reducing noise in the at least one selected windowed signal to obtain at least one clean signal;

[0035] -Step b) is included after the preprocessing stage:

[0036] - Apply a first bandpass filter, Hilbert transform, envelope analysis, and linear frequency modulated z-transform or numerical Fourier transform to the at least one clean signal to obtain at least one first power spectral density function.

[0037] - Identify the first spectral peak in the at least one first power spectral density function.

[0038] - Assess the subject's heart rate based on the first spectral peak;

[0039] -Step b) is included after the preprocessing stage:

[0040] - Apply a second bandpass filter and a linear frequency modulated z-transform or a digital Fourier transform to the at least one clean signal to obtain at least one second power spectral density function.

[0041] - Identify the second spectral peak in the at least one second power spectral density function.

[0042] - Assess the subject's respiratory rate based on the second spectral peak;

[0043] -When the measurement unit includes at least one accelerometer and at least one gyroscope, the at least one first spectral density function includes a first accelerometer power spectral density function derived from the measurement signal transmitted by the at least one accelerometer and a first gyroscope power spectral density function derived from the measurement signal transmitted by the at least one gyroscope.

[0044] - Identifying the first spectral peak includes identifying the accelerometer spectral peak in the first accelerometer power spectral density function and the gyroscope spectral peak in the first gyroscope power spectral density function, and assessing the subject's heart rate includes comparing a first relative band power of a selected accelerometer spectral peak with a second relative band power of a selected gyroscope peak.

[0045] - When the measurement unit includes at least one accelerometer and at least one gyroscope, the at least one accelerometer is configured to measure time acceleration signals along three orthogonal axes, and the at least one gyroscope is configured to measure time angular rotation signals along three orthogonal axes. Step b) includes performing principal component analysis on the time acceleration signals and the time angular rotation signals after a preprocessing phase to assess the subject's heart rate and respiratory rate.

[0046] -Step b) further includes:

[0047] - Process the at least one first power spectral density signal to obtain at least one cardiac resonator signal.

[0048] - Process the at least one echocardiogram signal to obtain at least one windowed aortic valve opening peak signal.

[0049] - Extract the local minimum and maximum values ​​of the at least one windowed aortic valve opening peak signal.

[0050] - Annotate the local minimum and maximum values ​​using reference points;

[0051] -Step b) further includes:

[0052] - Use the subject's respiratory rate to identify the respiratory cycle.

[0053] - Process the respiratory cycle to obtain the subject's inspiratory volume, an estimate of the subject's lung capacity, an estimate of the subject's vital capacity, and an estimate of the subject's inspiratory and expiratory phases. Detailed Implementation

[0054] The following description, given with reference to the accompanying drawings, will make the scope of the invention and the manner in which it is implemented clear. The invention is not limited to the embodiments shown in the drawings. Accordingly, it should be understood that where features mentioned in the claims are followed by reference numerals, such reference numerals are included only for the purpose of enhancing the comprehensibility of the claims and are in no way intended to limit the scope of the claims.

[0055] In the attached diagram:

[0056] - Figure 1 This demonstrates a comparison between cardiac plethysmography (SCG) measurements and electrocardiograms;

[0057] - Figure 2 A system for determining the cardiac or respiratory activity of a subject, according to the present invention, is illustrated schematically.

[0058] - Figure 3 This illustrates an eyeglass frame according to an embodiment of the present invention;

[0059] - Figure 4 This represents a measurement signal according to an embodiment of the present invention;

[0060] - Figure 5 The main steps of the method according to the present invention for determining the cardiac or respiratory activity of a subject are illustrated schematically.

[0061] - Figure 6 It indicates that, according to the present invention Figure 4 The preprocessed signal generated by the measurement signal processing steps;

[0062] - Figure 7 This indicates the product generated by the processing steps according to the invention. Figure 5 The envelope signal of the preprocessed signal;

[0063] - Figure 8 This indicates the product generated by the processing steps according to the invention. Figure 6 The power spectral density of the envelope signal;

[0064] - Figure 9An example of a power spectral density curve generated by the processing steps according to the present invention is shown;

[0065] - Figure 10 A scatter plot showing the measurement results obtained using the system according to the invention for determining the cardiac activity of a subject;

[0066] - Figure 11 This represents the measurement signals (raw and preprocessed) obtained using a system according to the invention designed to determine the respiratory activity of a subject;

[0067] - Figure 12 This indicates the product generated by the processing steps according to the invention. Figure 11 The power spectral density of the preprocessed measurement signal;

[0068] - Figure 13 A flowchart illustrating the steps of a method for characterizing the cardiac activity of a subject according to the present invention;

[0069] - Figure 14 This represents a typical cardiac resonant waveform with a reference point;

[0070] - Figure 15 This represents the detector signal according to the present invention;

[0071] - Figure 16 This indicates that the presentation is applied to Figure 15 The peak signal detected by the threshold method of the detector signal;

[0072] - Figure 17 This indicates additional peak values ​​obtained by supplementing the processing steps according to the invention. Figure 16 The signal;

[0073] - Figure 18 It indicates that by Figure 17 The aortic valve opening peak value was obtained through signal analysis;

[0074] - Figure 19 This indicates that the application according to the present invention... Figure 18 The signal processing steps yielded the echocardiogram waveform.

[0075] - Figure 20 It indicates Figure 19 Annotated version of the cardiac waveform;

[0076] - Figure 21 The flowchart illustrates the steps of a method for characterizing the respiratory activity of a subject according to the present invention.

[0077] - Figure 22 It is a representation of human lung capacity and vital capacity.

[0078] Figure 2 A system 1 for determining the cardiac or respiratory activity of a subject according to the present invention is schematically illustrated. System 1 includes an accessory 2 configured to be worn on the subject's head. Accessory 2 may be, for example, an eyeglass frame, headphones, earphones, or jewelry. Accessory 2 is provided with a measurement unit 3, which includes at least one accelerometer 31 or at least one gyroscope 32. Measurement unit 3 may also be an inertial measurement unit, comprising three accelerometers measuring linear acceleration values ​​along three orthogonal axes and three gyroscopes measuring rotational speeds around the three orthogonal axes.

[0079] At least one accelerometer 31, for example, a single-axis accelerometer, a dual-axis accelerometer, or a triaxial accelerometer. At least one gyroscope 32, for example, a single-axis gyroscope, a dual-axis gyroscope, or a triaxial gyroscope.

[0080] Measurement unit 3 transmits at least one measurement signal that reports the kinematic characteristics or position of Annex 2 and thus the kinematic characteristics or position of the subject's head. The kinematic characteristics may be velocity in one direction, rotational speed about one direction, or acceleration in one direction.

[0081] Annex 2 also includes processing unit 4. Processing unit 4 is programmed to perform the following steps:

[0082] - Receive and record at least one measurement signal from measurement unit 3 for at least a period of time to obtain sample measurement results.

[0083] - Process the sample measurements to form at least one recorded signal for use in determining the cardiac or respiratory activity.

[0084] The steps for processing sample measurement results can be implemented in an embedded manner, that is, autonomously as described in Annex 2. Alternatively, the processing steps can be implemented remotely on a remote server receiving the sample measurement results.

[0085] Figure 3 An illustration is given of an accessory 2 consisting of an eyeglass frame, including a measuring unit 3 (not shown). The measuring unit 3 can be located anywhere on the eyeglass frame, such as in one of the temples, in the front of the frame, or in one of the tips of the frame. In this particular case, when the eyeglasses are a pair of connected glasses, the measuring unit 3 can advantageously be a sensor already embedded in the connected glasses, such as an inertial sensor for measuring body activity, typically the accelerometer of a pedometer. Figure 3 In the diagram, three orthogonal linear axes x, y, and z are represented. Measurement unit 3 can then be a complete inertial measurement unit that measures the linear acceleration values ​​along these three axes and the rotational speed around these three axes.

[0086] Figure 4 Typical raw signals obtained using measurement unit 3 (here along) Figure 3 The diagram shows the accelerometer signal along the z-axis.

[0087] Therefore, as will be shown below, the system 1 according to the invention allows for the acquisition of measurements such as cycloplegic resonance or gyrocardiography simply by wearing the attachment on the head in a permanent or semi-permanent manner, without having to position any sensors near the heart and without the subject being in a given location.

[0088] The purpose of processing the sample measurement results obtained from measurement unit 3 of system 1 is to implement a method for determining the cardiac or respiratory activity of a subject wearing accessory 2 on their head. Cardiac activity refers to heart rate and quantitative and qualitative descriptions of the subject's heartbeat. Respiratory activity refers to respiratory rate and quantities such as the subject's inhaled volume and lung capacity.

[0089] Figure 5 The main steps of the method for determining the cardiac or respiratory activity of a subject wearing the head-mounted system 1 in an accessory 2 are described.

[0090] The determination method begins with triggering step S0, which is designed to trigger data acquisition by measurement unit 3.

[0091] During the triggering step S0, it is first determined in sub-step S01 whether the subject is wearing accessory 2. This can be done manually (the subject knows they are wearing the accessory) or by detecting the subject's activity level. For example, if the signal measured by measurement unit 3 is sufficiently stable or repetitive, i.e., there is no obvious activity, it can be inferred that the subject is, for example, resting or walking, and thus a measurement result can be obtained.

[0092] In the case of an eyeglass frame in Annex 2, the indication that the signal measured by the measuring unit 3 is a monotonic signal can be supplemented by measuring the orientation of the eyeglass frame worn by the subject. For example, using Figure 3 The acceleration of the stage along the z-axis, observed along the axis, may constitute this indication.

[0093] Then, in sub-step S02, the sampling rate and sensitivity of measurement unit 3 are increased. If measurement unit 3 is an inertial measurement unit, the accelerometer range is set to, for example, 2g, with a resolution of 0.122mg, and the gyroscope range is set to 2000 degrees per second, with a resolution of 16.4 degrees per second. A more accurate and sensitive inertial measurement unit can be used as measurement unit 3. The firmware embedded in the inertial measurement unit increases the sampling frequency from 5Hz to, for example, 200Hz, or, for example, to a sampling frequency between 100Hz and 200Hz.

[0094] Then, in sub-step S03, data acquisition by measurement unit 3 is triggered.

[0095] As a variation, the triggering step S0 can be performed manually. For example, data acquisition by the measurement unit 3 can be triggered by the subject's request to activate a switch button or by a remote request sent by the subject's mobile phone or any other wearable device, thereby activating the measurement unit 3.

[0096] Following triggering step S0 is acquisition step S1, during which at least one measurement signal is recorded in a time slot. To determine the subject's cardiac activity, measurement signals are recorded, for example, in time slots of at least 10 seconds. To determine the subject's respiratory activity, measurement signals are recorded, for example, in time slots of at least 60 seconds.

[0097] Following acquisition step S1 is preprocessing step S2, which aims to assess the quality of the recorded signal in order to efficiently extract features of interest, namely, features related to the subject's cardiac and respiratory activities.

[0098] The values ​​constituting the recorded signal are timestamped, with units such as ticks or milliseconds. In sub-step S21 of preprocessing step S2, the timestamp unit is converted to seconds. After the time conversion in sub-step S21, the recorded signal is downsampled in sub-step S22 by interpolation (e.g., by linear interpolation, cubic interpolation, or polynomial interpolation). For example, to balance processing time and accuracy, a downsampling frequency of 100 Hz can be used. This value is suitable for analyzing cardiac and respiratory activities where important signal features exhibit frequencies below 50 Hz.

[0099] Following the downsampling sub-step S22, sub-step S23 is executed, which involves searching for peaks in the recorded signal. A thresholding algorithm is used to perform the peak search.

[0100] For example, a search for peaks detects peaks with a salience of 0.2. Peak salience measures the degree to which a peak stands out due to its inherent height and its position relative to other peaks. To measure the salience of a peak P, the following procedure can be followed: place a mark on peak P; extend horizontal lines from peak P to the left and right until the lines cross the signal or reach the left or right end of the signal due to a higher peak; find the minimum value of the signal in each of the two defined intervals (left and right), which is either a valley or one of the signal endpoints; specify a reference level L for the higher of the two interval minimums. ref At this reference level L ref The height of the peak P above is its prominence.

[0101] If no peak is found in sub-step S23, meaning the recorded signal contains only noise, the method returns to acquisition step S1 to acquire a signal of better quality.

[0102] If a peak is detected during sub-step S23, sub-step S24 is executed, which includes noise reduction in the recorded signal. Noise reduction in sub-step S24 is performed within a so-called "optimal time window." These optimal time windows are portions of the recorded signal with a predetermined duration selected from the time slots in which the recorded signal is recorded. An optimal time window is one where most peaks have globally similar amplitudes and artifacts in the recorded signal are minimized. Figure 6 An example of a preprocessed signal within a selected optimal time window is shown. The signal is a time signal a. z (t), where t represents time, is measured by the accelerometer 31 of the measuring unit 3.

[0103] Then, once the optimal time window is selected, the recorded signal (denoted as S) is analyzed. BEST (t) Apply filters (e.g., 3rd or 5th order moving average filters, or bandpass filters, or, for example, a combination of low-pass and high-pass Butterworth filters with corresponding cutoff frequencies of 0.1 Hz and 50 Hz) to reduce noise on the corresponding portions of the recorded signal.

[0104] The noise reduction filter parameters change when considering cardiac activity estimation or respiratory activity estimation.

[0105] For example, bandpass filters with cutoff frequencies of 4Hz and 50Hz can be applied to signal S. BEST The Fast Fourier Transform (FFT) of (t) is used for cardiac activity estimation, while bandpass filters with cutoff frequencies of 0.1 Hz and 0.9 Hz can be applied for respiratory activity estimation. In practice, the bandwidth of the filter applied for heart rate estimation is typically configured to focus on the heart rate range between 40 heartbeats per minute and 240 heartbeats per minute, while the bandwidth of the filter applied for respiratory rate estimation is configured to focus on the respiratory rate range between 6 breaths per minute and 40 breaths per minute.

[0106] By using moving average filters, polynomial smoothing, Savitzky-Golay filters, median filters, comb filters, and empirical mode decomposition, artifacts caused by involuntary or voluntary movements of the subject's body (such as walking, talking, yawning) can be further removed.

[0107] When the noise reduction sub-step S24 is executed, the preprocessing step S2 ends.

[0108] Then, the rate estimation step S3 (heart rate R) is performed after the sub-steps described below. Hand / or respiratory rate R R ).

[0109] Here, we assume a signal (accelerometer signal a) corresponding to the signal obtained after performing preprocessing step S2. BEST (t) or gyroscope signal g BEST (t) is available (t is time in seconds).

[0110] In estimating the subject's respiratory rate R R In this case, the following sub-steps S31 and S32 are optional.

[0111] In sub-step S31, signal a is calculated. BEST (t)(correspondingly g) BEST The Hilbert transform of the time function h(t). The Hilbert transform of the time function h(t) is defined by the following equation:

[0112]

[0113] in and

[0114] The output of sub-step S31 is a signal. (correspondingly) ).

[0115] In substep S32, the Hilbert transform is then calculated. Accordingly Envelope analysis, thereby outputting signal A BEST (t), correspondingly G BEST (t). Figure 7 The solid line shows the... Figure 6 signal a BEST (t)(dashed line) Signal A obtained after implementing sub-steps S31 and S32. BEST (t).

[0116] In sub-step S33, for signal A BEST (t), correspondingly G BEST (t)(or if sub-steps S31 and S32 are not executed, then for signal a) BEST (t) or g BEST (t) Apply the linear frequency modulated z-transform or, for example, the discrete Fourier transform (DFT) implemented by the fast Fourier transform (FFT) to output the signal F(A)(f), and correspondingly F(G)(f), where f represents the frequency in Hz.

[0117] The expression for the linear frequency modulated z-transform of the sampled signal h(n), n = 0..N-1, is given below:

[0118] For k = 0, 1, ..., M-1,

[0119] And for k = 0, 1, ..., M-1, z k =A*W -k A represents the starting point of the complex number, and W is the point z. j The complex ratio between them, and M is the number of points.

[0120] The expression for the digital Fourier transform of a sampled signal h(n) including N samples (n = 0, 1, ..., N-1) is given below:

[0121] For k = 0, 1, ..., N,

[0122] In sub-step S34, the power spectral density DSPA(f) and the corresponding DSPG(f) are calculated based on the signal F(A)(f) and the corresponding F(G)(f):

[0123]

[0124]

[0125] Where T is the average time interval.

[0126] Other ways to calculate the power spectral density DSPA(f) and correspondingly DSPG(f) can be to calculate the autocorrelation of the signal F(A)(f) and correspondingly F(G)(f), or to use Welch's method.

[0127] Figure 8 An example of the power spectral density curve obtained after performing sub-step S34 is given, where the frequency corresponding to the maximum power spectral density is indicated by a black circle (approximately 1.7 Hz).

[0128] In sub-step S35, the frequency f with the maximum power spectral density in the signal DSPA(f) and correspondingly DSPG(f) is first identified. max Then, the heart rate R is estimated by identifying the fundamental frequency associated with higher harmonics as follows. H Corresponding respiratory rate R R .

[0129] In fact, sometimes the first harmonic of a signal may have a higher power spectral density than the fundamental frequency. The solution used in this invention includes considering the power spectral density at frequencies f. max / 2 and f max The power spectral density curve centered on the center consists of two parts, P1 and P2, and each part is presented with a width of f. maxThe bandwidth is 2 / 2. Then, the ratio between the area under the power spectral density curve in portion P1 and the area under the power spectral density curve in portion P2 is calculated. As implemented in this disclosure, this ratio is compared to a value of 0.01. If the ratio is less than 0.01, the rate to be estimated (heart rate R) is inferred. H Corresponding respiratory rate R R ) corresponds to frequency f max If the ratio is higher than 0.01, then an adjustment is performed on frequency f. max Search for the peak of the power spectral density curve near / 2. If a peak is found close to frequency f... max / 2, then it is inferred that the rate to be estimated corresponds to the frequency f. max / 2.

[0130] Figure 9 This method of identification is demonstrated: it can be seen that the power spectral density curve at f max The maximum peak value appears at approximately 2.86 Hz. Then, following the method described above, consider two frequency ranges: each with a frequency of f... max / 2 and f max A bandwidth centered at 1.43 Hz was observed. An area ratio greater than 0.01 was found, and a peak centered at 1.46 Hz was detected in the first frequency range, indicated by a black circle, very close to frequency f. max / 2. Therefore, the final estimate is 1.46 Hz. To evaluate the searched rate, the following formula is used:

[0131] R H (bps)=f max (Hz) and R H (bpm) = 60f max ,

[0132] bps is an abbreviation for heart rate per second, and bpm is an abbreviation for heart rate per minute.

[0133] In the previous example, a frequency of 1.46 Hz corresponds to a heart rate of 88 bpm.

[0134] After executing sub-step S35, the rate estimation step S3 ends.

[0135] Table 1 below shows a series of results obtained from the estimation of several signal samples. These results compare reference values ​​of heart rate with heart rate values ​​obtained using the method according to the invention and with system 1.

[0136] Table 1

[0137]

[0138] The mean error was 2.69 bpm, and the standard deviation was 4.09 bpm.

[0139] Figure 10 A scatter plot corresponding to these measurements is shown, fitted with linear regression, presenting the R² values. 2 The correlation coefficient is 0.97.

[0140] In a preferred embodiment, when the measuring unit 3 includes at least a measuring edge Figure 3 The z-axis signal a z The acceleration timing of (t) is given by signal a. z (t) is used as the recorded signal processed in steps S0, S1, S2 and S3.

[0141] Figure 11 and Figure 12 The following is given for estimating the respiratory rate R of the subject. R A diagram of the signal.

[0142] In a preferred embodiment, in order to estimate the subject's heart rate R H The accelerometer 31 from the measurement unit 3 specifically measures along the... Figure 3 The signal of the acceleration value along the z-axis is used to implement steps S0 to S3.

[0143] Figure 11 An example is shown of the time signal used to determine the respiratory rate estimate and its denoised version (a smooth black curve with a white border) after applying sub-step S24. Here, more precisely, bandpass filters with cutoff frequencies of 0.1 Hz and 0.9 Hz are applied. Figure 11 The blue curve.

[0144] Figure 12 The corresponding power spectral density curve obtained after performing the power spectral density calculation sub-step S34 is shown. In this graph, the frequency corresponding to the maximum power spectral density, here 0.25 Hz, as indicated by the black circle, can be evaluated. The corresponding respiratory rate at this time is 15 breaths per minute.

[0145] In the embodiment, when estimating heart rate R H In cases where both accelerometer and gyroscope signals are available, these two signals can be combined to increase the accuracy of rate estimation.

[0146] The combination is performed according to the following decision algorithm.

[0147] If the frequency f obtained from the accelerometer signal analysis a and the frequency f obtained from the gyroscope signal analysis g If the values ​​are close, then the average value is chosen as the search rate.

[0148]

[0149] If the frequency f obtained from the accelerometer signal analysis a The peak value at that point exhibits a relative power band power higher than 0.9, and the frequency f obtained from the gyroscope signal analysis... g If the peak value at a certain point exhibits a relative band power of less than 0.9, then a frequency f is chosen. a The corresponding heart rate R is measured in heartbeats per second. H As the rate found. More specifically, frequency f a Given in Hz:

[0150] R H (bps)=f a And R H (bpm) = 60f a

[0151] Frequency f a Peak or frequency f at the point g The "relative band power" of the peak value at a given frequency f refers to the power relative to the frequency f. a The power in the frequency band corresponding to the peak value at that point (calculated as a percentage of the total signal power), and correspondingly related to the frequency f. g The power in the frequency band corresponding to the peak value at that point. Based on frequency f a Correspondingly f g The peak bandwidth at that location is used to calculate the frequency band.

[0152] If the frequency f obtained from the accelerometer signal analysis a The peak value at that point exhibits a relative power of less than 0.9, and the frequency f obtained from the gyroscope signal analysis... g If the peak value at a certain point exhibits a relative band power higher than 0.9, then a frequency f is chosen. g The corresponding heart rate R is measured in heartbeats per second. H As the rate at which the search is conducted.

[0153] R H (bps)=f g And R H (bpm) = 60f g

[0154] Otherwise, if a heart rate estimate has just been performed, then choose f. a and f g The closest heart rate R to the previous estimate H The frequency.

[0155] Otherwise, if the frequency f obtained from the gyroscope signal analysis gThe peak value at that point exhibits a frequency f that is higher than that obtained from the analysis of the accelerometer signal. a A higher relative band power at a peak value, for example, three times higher, means that the clarity of the estimate from the gyroscope is three times higher than the clarity of the estimate from the accelerometer. Therefore, choosing a frequency f... q The corresponding rate is used as the rate found in the search.

[0156] Otherwise, choose the frequency f obtained from the accelerometer signal analysis. a The corresponding rate is used as the rate found in the search.

[0157] In the embodiment, when the three accelerometer signals a x (t), a y (t) and a z (t) and three gyroscope signals g x (t), g y (t) and g z (t) is available when (where x, y, and z are, for example, with) Figure 3 The axes (represented in the text) can be combined to increase heart rate R. H Accuracy or respiratory rate R R The accuracy of the estimate.

[0158] Assume signal a on one side x (t), a y (t), a z (t) and the other signal g x (t), g y (t), g z (t) is the signal obtained after sub-step S23, and peaks were found in these signals.

[0159] They are combined using principal component analysis (PCA) to determine the best direction for capturing the maximum variance.

[0160] The principle of PCA is to fit a p-dimensional ellipsoid to the data, where the data is one aspect of the signal a. x (t), a y (t), a z (t) and the other signal g x (t), g y (t), g z (t). Each of the p axes of the elliptic is characterized by a variation value (i.e., the total distance between projected data) and a variance value (i.e., the total distance from the original data to its corresponding projected data on the axis). The axis with the largest variance corresponds to the representation closest to the original data.

[0161] The lists of measured vectors (ax(t), ay(t), az(t)) and (gx(t), gy(t), gz(t)) are projected onto the optimal orientation axes found through principal component analysis, so that each triple a is represented as a vector. principal (t) and g principal (t) Outputs a single signal.

[0162] Then, by analyzing signal a principal (t) and g principal (t) After applying a 5th-order moving average filter, a combination of an ideal bandpass filter or a low-pass filter with a cutoff frequency of 0.1 Hz and a high-pass filter with a cutoff frequency of 0.5 Hz is used to perform the noise reduction sub-step S24. Sub-step S24 outputs two noise-reduced signals S. a (t)(from accelerometer measurements) and S g (t)(from gyroscope measurement results).

[0163] In this embodiment, the execution of the respiratory rate estimation step S3 differs from the previously described series of sub-steps S31 to S35, as follows. For signal S... a (t) and S g Each application in (t) has four sub-steps S31b to S34b.

[0164] In sub-step S31b, the denoised signal S is calculated. a (t), correspondingly S g The signal-to-noise ratio (SNR) of (t) a Correspondingly, SNR g .

[0165] In sub-step S32b, these signals F(S) are used. a (f) Correspondingly, F(S) g The signal S is computed using the linear frequency modulated z-transform of (f) or, for example, the discrete Fourier transform implemented by the Fast Fourier Transform (FFT). a Correspondingly S g Power spectral density:

[0166]

[0167]

[0168] In substep S33b, the power spectral density values ​​are used as weights to identify peaks and sort them in descending order. Any peak search method can be used, such as correlation, filtering, demodulation, or moving average.

[0169] In substep S34b, a larger weighted value is selected as the respiratory rate frequency f. Ra Correspondingly fRg .

[0170] In the final sub-step S35, the result f corresponding to the highest signal-to-noise ratio calculated at sub-step S31b is selected. R In the two results f Ra with f Rg Distinguishing between them.

[0171] Similar to heart rate estimation, respiratory rate is given by the following formula:

[0172] R R (bps)=f R (Hz)

[0173] R R (bpm) = 60 * f R (Hz)

[0174] In a preferred embodiment, in order to estimate the subject's respiratory rate R R Steps S0 to S3 are performed using signals from the gyroscope 32 of the measurement unit 3.

[0175] Once the subject's heart rate R is estimated at the end of step S3... H and / or the subject's respiratory rate R R This allows us to extract specific features of interest to characterize cardiac and respiratory activities.

[0176] The following describes the estimation of the subject's heart rate R. H and respiratory rate R R Then, two separate steps, S41 and S42, are used to characterize the subject's cardiac and respiratory activities, respectively.

[0177] First, step S41, used to characterize the subject's cardiac activity, will be described.

[0178] Figure 13 The flowchart of step 41 is shown schematically and will be described below.

[0179] Figure 14 The image shows a typical electrocardiogram (ECG) with points of interest between two heartbeats. Both systolic and diastolic points can be recorded on this typical ECG. These points of interest cannot be extracted from a conventional ECG. Systolic points include aortic opening (AO), peak ventricular ejection (PE), and aortic closure (AC). Diastolic points include mitral opening (MO), rapid ventricular filling (RF), atrial contraction (AS), and mitral closure (MC).

[0180] Step S41 aims to index those points on the signal that are typically measured by measurement unit 3.

[0181] Here, it is assumed that the accelerometer signal a is the same as the previously defined accelerometer signal a BEST (t) or gyroscope signal g BEST A corresponding time signal SCG(t) is available (t is time in seconds) in SCG(t), and is obtained after step S2. For example, the signal SCG(t) may have passed through bandpass filters with cutoff frequencies of 4Hz and 50Hz. It is also assumed that the heart rate R in units of heartbeats per second (bps) has been estimated after step S3. H Heart rate R H This corresponds to the rate between the two aortic valve opening peaks (AO) in the SCG(t) signal.

[0182] In the first sub-step S411, the detector signal x, defined by the following formula, is calculated. det (t) is used to detect the AO peak value:

[0183]

[0184] Optionally, in the detector signal x det In the formula for (t), the detector signal x can be calculated by considering only the negative part of SCG(t) or only the positive part of the SCG(t) signal. det (t).

[0185] Figure 15 It showed x det (t) An example of the shape of the signal, whose maximum value is indicated by a red dot.

[0186] Alternatively, this can be achieved by adjusting the signal xd. et (t) Apply a (e.g., 5th order) moving average filter to smooth the signal.

[0187] In the next sub-step S412, x is used det (t) The time index of the maximum value of the signal and the estimated heart rate R in bps. H The signal is divided into segments. More specifically, signal x det (t) is divided into durations of Dt = 1 / R H And it begins at time t max The segment is -0.215Dt, where t max x represents det (t) is the time index of the maximum value of the signal.

[0188] In the next sub-step S413, the local maximum value of each segment is stored.

[0189] In the next sub-step S414, a thresholding operation is performed at the 60th percentile, resulting in an output peak amplitude that is 60% higher than the amplitude of each local maximum. Figure 16 The signal x before and after threshold processing is shown. det The superposition of (t) versions. The peak value obtained after the thresholding operation is represented by a dotted curve.

[0190] In the next sub-step S415, the threshold signal x is calculated. det The interval between all consecutive peak pairs in (t).

[0191] Then, the following estimation algorithm is applied to each calculated interval I.

[0192] When I is less than 1 / R Hmax In the case of R Hmax This indicates the maximum permissible heart rate (usually R). Hmax (equal to 200 bpm): The peak values ​​are too close, so only the peak with the maximum value is considered. This allows for the detection of abnormal peak values.

[0193] When I is greater than 1.2*Dt, it can be inferred that peaks are lost in the corresponding intervals. The loss of peaks is obtained under the assumption that they are regularly distributed within interval I. Therefore, the loss of peaks is theoretically estimated. To aid in the estimation, measured peaks that approximate the theoretically estimated peaks are used.

[0194] Figure 17 Signal x was shown det The detected peak (represented by a black hollow circle) in (t) has already been included in the peak (represented by a black solid circle) obtained after the threshold operation in sub-step S414.

[0195] Otherwise, the interval is considered acceptable, and two corresponding peaks (i.e., the peaks defining the interval) are considered and retained for analysis.

[0196] Figure 18 The diagram illustrates the correspondence between the peak values ​​detected after implementing the aforementioned estimation algorithm and the AO peak value. More specifically, the AO peak value corresponds to the signal x. det Each detected peak in (t) is associated with a specific peak.

[0197] It can be noted that the AO peak can be used to assess a subject's heart rate variability (HRV). HRV is a physiological phenomenon of variation in the time interval between heartbeats. It is measured by changes in the heartbeat interval. HRV is a good indicator of both cardiovascular and non-cardiovascular diseases.

[0198] In sub-step S416, the signal SCG(t) is segmented. The median curve SCG is calculated for all segments.median (t), which normalizes the amplitude of the signal in each segment. Figure 19 This is an example of the median curve obtained as a result of sub-step S416. On this curve, the point with the x-coordinate of 0 corresponds to the median of all AO peaks.

[0199] In sub-step S417, the median curve SCG is positioned as follows: median (t) identifies the baseline point for typical cardiac resonance recordings. From the intermediate curve SCG median Extract duration Dt from (t) and start at approximately t. AO A window of -0.25Dt, t AO The time index represents the AO peak value. Local minimum and maximum values ​​are then extracted to annotate the signal. Figure 20 The median curve with annotations is shown.

[0200] Typically, the IM-AO-IC-RE sequence is used to identify cases where the amplitude during diastole is greater than the amplitude during systole.

[0201] As a variation, more than one signal S can be used. BEST (t), that is, in order to smooth the median curve and increase the signal-to-noise ratio, several optimal windows can be considered. For example, for each new optimal window, the signal S is acquired. BEST_i (t) can be used to calculate the associated heart rate R. Hi Then, it can be based on R. Hi The values ​​are time-scaled across all signals to obtain a common reference.

[0202] As another variation, machine learning algorithms, such as neural networks, can be used to perform benchmark identification.

[0203] Now, step S42, used to characterize the subject's respiratory activity, will be described.

[0204] Figure 21 The flowchart for step 42 is shown schematically and will be described below.

[0205] Here, it is assumed that the accelerometer signal a is the same as the previously defined accelerometer signal a BEST (t) or gyroscope signal g BEST A corresponding time signal S in (t) R (t) is available (t is time in seconds) and was obtained after performing preprocessing step S2. For example, the signal SCG(t) may have undergone bandpass filters with cutoff frequencies of 0.1 Hz and 0.9 Hz. It is also assumed that the respiratory rate R, in heart rate per second (bps), has been estimated. R .

[0206] Typical quantities of interest are a subject's inspiratory volume and vital capacity. The air in the lungs is measured based on lung capacity and vital capacity. Volume measures the amount of air in a function (such as inhalation or exhalation). Volume is any two or more volumes. An example of volume is how much air can be inhaled starting from the end of a maximal exhalation.

[0207] Figure 22 This is a diagram illustrating human lung capacity and vital capacity. Total vital capacity (TLC) is the volume of air contained in the lungs at the end of a maximal inspiration. Inspiratory volume (IC) is the maximum volume of air that can be inhaled after a normal quiet exhalation. Inspiratory reserve (IRV) is the additional volume of air that can be inhaled with maximal effort after a normal quiet inspiration. Tidal volume (TV) is the volume of air inhaled or exhaled during each normal quiet respiratory cycle. IC is the sum of IRV and TV. Functional residual volume (FRC) is the volume of air remaining in the lungs at the end of a normal quiet exhalation. FRC is the sum of ERV and RV. Expiratory reserve (ERV) is the additional volume of air that can be exhaled with maximal effort beyond the level reached at the end of a normal quiet exhalation. Residual volume (RV) is the volume of air remaining in the lungs at the end of a maximal exhalation. Figure 22 This demonstrates how these different volumes and cavities can be obtained by analyzing time signals representing respiration over a period of time.

[0208] In the first sub-step S421, in the time signal S R The respiratory cycle is identified in (t). More specifically, in the time signal S R The fitting time period in (t) is 1 / R R The sine curve is fitted. The inhaled volume is estimated by combining the amplitude of the fitted sine curve with an estimate of the subject's lung capacity based on the subject's height.

[0209] In the next sub-step S422, the subject's vital capacity is estimated as follows: First, a Hilbert transform of an identified respiratory cycle is calculated. The subject's vital capacity is estimated as the average of the analytical envelope of the previously calculated Hilbert transform. In a variant, to improve estimation accuracy, a weighted average of different values ​​obtained for different respiratory cycles can be used.

[0210] In the next sub-step S423, the respiratory phases, namely the inhalation phase and the exhalation phase, can be identified.

[0211] By analyzing the time signal S R The instantaneous phase assessment in the Hilbert transform of the lowest frequency component of (t) is combined to predict the inspiratory phase. The instantaneous phase is then correlated with the phase of the peak frequency of the respiratory power spectral density DSPG(f).

[0212] The phase of the sine curve fitted at substep S421 provides information about the subject’s respiratory state, i.e., whether it is inhalation or exhalation, regardless of the point in the respiratory cycle.

[0213] Therefore, System 1 according to the invention allows for the estimation of cardiac or respiratory activity in a subject wearing Accessory 2 in a permanent or semi-permanent manner, without any constraints on the position of the accessory or the subject itself. The solution proposed by the invention is cost-effective, low-power, and space-saving.

[0214] Advantageously, while classical SCG techniques, for example, require sensors to be close to the chest to obtain a clear signal, the system 1 according to the invention is able to extract an estimate of the subject's cardiac or respiratory activity based on measurements performed away from the heart and chest.

[0215] Since System 1 is worn permanently or semi-permanently via Annex 2, an appropriate time slot can be selected to perform measurement result acquisition.

[0216] Finally, the measurement unit 3, which provides signals for estimating cardiac and / or respiratory activity, is not limited to this application but can be shared with other functions in Annex 2, such as step counting, posture monitoring, power management, etc.

[0217] Variation:

[0218] The processing performed in steps S2, S3, S41, and S42 can be applied to any signal or any combination of signals from the accelerometer 31 and / or gyroscope 32 of the measurement unit 3. For example, steps S2, S3, S41, and S42 can be applied to signals with minimal noise, or when an inertial measurement unit comprising three accelerometers and three gyroscopes is used as the measurement unit 3, it can be applied to vectors (a x a y a z ) or (g x g y g z The amplitude of signal a is applied to signal a. x a y and a z and g x g y and g z The weighted average, or any linear or nonlinear combination applied to these signals.

[0219] Other signal processing methods, such as directional peak detection, polynomial spline fitting, wavelet transform, and empirical mode decomposition, can be used to remove noise and artifacts related to the subject's body motion, or to improve heart rate accuracy. H or respiratory rate R RThe estimation replaces the calculation of the Fourier transform.

Claims

1. A method for determining the cardiac or respiratory activity of a subject, the method comprising: The steps of measuring and recording sample measurements related to the kinematic characteristics or position of at least the attachment (2) worn on the head of the subject over a period of time, and the steps of processing these sample measurements to determine the cardiac or respiratory activity. The characteristic feature is that the processing steps include a preprocessing stage with the following steps: - Convert the timestamps of the sample measurement results into the selected time unit. - The sampling frequency of the recorded sample measurement results is downsampled to obtain at least one downsampled signal. - Apply at least one window function to the at least one downsampled signal to obtain at least one windowed signal. - Select at least one windowed signal from the at least one windowed signal. - The quality of the at least one selected windowed signal is estimated by searching for peaks in the at least one selected windowed signal.

2. The method according to claim 1, wherein, The attachment (2) is equipped with at least one motion sensor.

3. The method according to claim 1, wherein, The processing steps are executed in an embedded manner or on a remote host.

4. A system (1) for determining the cardiac or respiratory activity of a subject, the system (1) comprising an accessory (2) configured to be worn on the head of the subject, and the accessory being provided with: a measurement unit (3) including at least one motion sensor and transmitting at least one measurement signal reporting at least the position, velocity, rotational speed, or acceleration of the accessory (2); and a processing unit (4) programmed to perform the following steps: - a) Receive and record at least one measurement signal from the measurement unit (3) for at least a period of time to obtain sample measurement results. - b) Process the sample measurements to form at least one recorded signal for determining the cardiac or respiratory activity. Its features are, Step b) includes a preprocessing stage with the following steps: - Convert the timestamps of the sample measurement results into the selected time unit. - The sampling frequency of the at least one recorded signal is downsampled to obtain at least one downsampled signal. - Apply at least one window function to the at least one downsampled signal to obtain at least one windowed signal. - Select at least one windowed signal from the at least one windowed signal. - The quality of the at least one selected windowed signal is estimated by searching for peaks in the at least one selected windowed signal.

5. The system (1) according to claim 4, wherein, The accessory (2) is selected from the following: eyeglass frames, eyeglass attachments, eyeglass clips, eyeglass frames with straps, headphones, earphones, and jewelry.

6. The system (1) according to claim 4, wherein, The processing unit (4) is further programmed to perform the following steps before recording the at least one measurement signal: - Determine whether the sample measurement results from the measurement unit (3) meet the triggering conditions. - If the triggering condition is met, increase the sampling rate, sampling duration, and sensitivity of the measurement unit (3). - Trigger the recording of the at least one measurement signal. Furthermore, the recording of the at least one measurement signal is performed over a time slot included within the at least one period of time for obtaining the sample measurement results that form at least one recorded signal.

7. The system (1) according to claim 4, wherein, When a peak is detected during the quality estimation step, the preprocessing stage further includes a step of reducing noise in the at least one selected windowed signal to obtain at least one clean signal.

8. The system (1) according to claim 7, wherein, Step b) is included after the preprocessing stage: - Apply a first bandpass filter, Hilbert transform, envelope analysis, and linear frequency modulated z-transform or numerical Fourier transform to the at least one clean signal to obtain at least one first power spectral density function. - Identify the first spectral peak in the at least one first power spectral density function. - assessing the heart rate (R H ).

9. The system (1) according to claim 7, wherein, Step b) is included after the preprocessing stage: - Apply a second bandpass filter and a linear frequency modulated z-transform or a digital Fourier transform to the at least one clean signal to obtain at least one second power spectral density function. - Identify the second spectral peak in the at least one second power spectral density function. - assessing the respiratory rate (R R ).

10. The system (1) according to claim 8, wherein, The measurement unit (3) includes at least one accelerometer (31) and at least one gyroscope (32), wherein the at least one first power spectral density function includes a first accelerometer power spectral density function (DSPA1) derived from the measurement signal transmitted by the at least one accelerometer (31) and a first gyroscope power spectral density function (DSPG1) derived from the measurement signal transmitted by the at least one gyroscope (32).

11. The system (1) according to claim 10, wherein: - Identifying the first spectral peak includes identifying the accelerometer spectral peak in the first accelerometer power spectral density function (DSPA1) and the gyroscope spectral peak in the first gyroscope power spectral density function (DSPG1). - assessing the heart rate (R H ) of the subject comprises comparing a first relative band power of a selected accelerometer spectral peak with a second relative band power of a selected gyroscope spectral peak.

12. The system (1) according to claim 7, wherein, The measurement unit (3) includes at least one accelerometer (31) and at least one gyroscope (32), the at least one accelerometer (31) being configured to measure the time acceleration signal a along three orthogonal axes. x (t), a y (t) and a z (t), and the at least one gyroscope (32) is configured to measure the time angular rotational speed signal g along three orthogonal axes. x (t), g y (t) and g z (t), where step b) includes processing the time acceleration signal a after the preprocessing stage. x (t), a y (t) and a z (t) and the time angle rotation signal g x (t), g y (t) and g z (t) Principal component analysis was performed to assess the heart rate (R) of the subjects. H ) and the respiratory rate (R) of the subjects. R ).

13. The system (1) according to claim 8, wherein, Step b) further includes: - Process the at least one first power spectral density function to obtain at least one cardiac plethysmography (SCG) signal. - Process the at least one cardiac plethysmography (SCG) signal to obtain at least one windowed aortic valve opening (AO) peak signal. - Extract the local minimum and maximum values ​​of the at least one windowed aortic valve opening (AO) peak signal. - Use benchmarks to annotate the local minimum and maximum values.

14. The system according to claim 9, wherein, Step b) further includes: - Using the subject's respiratory rate (R) R Identify the respiratory cycle. - Process the respiratory cycle to obtain the subject's inspiratory volume, an estimate of the subject's lung volume, an estimate of the subject's vital capacity, and an estimate of the subject's inspiratory and expiratory phases.

Images