Detection of passenger breathing

Abstract

The invention relates to a method implemented by a UWB system (30) in a vehicle for detecting the breathing of one or more passengers. The UWB system (30) comprises one or more UWB sensors (31). The method comprises, for each UWB sensor (31), the acquisition (S10) by the sensor of a radar echo signal that has an acquisition sampling frequency and a plurality of breathing search processes (S20). The breathing search processes (S20) comprise a first process (S21) and a second process (S22). The method improves the detection of breathing of one or more passengers.

Description

Passenger Breath Detection

[0001] The present disclosure relates to the detection of breathing of one or more passengers, and more particularly to a method implemented by a UWB (acronym for "Ultra Wide Band") system for the detection of breathing of one or more passengers, a computer program for a UWB system, a UWB sensor of a UWB system and a UWB system. Technical background

[0002] There are currently vehicles equipped with a UWB system comprising one or more UWB sensors installed in the vehicle. Such a UWB system may be capable of detecting the breathing of passenger(s). To do this, the UWB system may execute, for each UWB sensor, a breathing search process from an echo-radar signal acquired by the UWB sensor at an acquisition sampling frequency. In particular, this search process generally takes as input a sub-sample of a portion of the signal having a predetermined duration. The sub-sample may in particular be carried out at a frequency less than or equal to the acquisition sampling frequency.

[0003] However, performing such a search process is not efficient. Indeed, the breathing rate of a passenger can vary depending on the type of passenger, or their age. A child, for example, has a higher average breathing rate than an adult, or an animal has a higher average breathing rate than a human. Thus, to cover a sufficient frequency range, the subsample must be fine enough to detect high breathing frequencies and of a predetermined duration long enough to detect low breathing frequencies. This results in a large number of measurements to be stored in memory and analyzed by the search process. Executing the search process is therefore inefficient and the amount of memory used is significant.

[0004] Therefore, there is a need to improve the detection of breathing of one or more passengers. Summary

[0005] To this end, a method is proposed implemented by a UWB system in a vehicle for detecting the breathing of one or more passengers. The UWB system comprises one or more UWB sensors. The method comprises, for each UWB sensor, an acquisition by the sensor of an echo-radar signal having an acquisition sampling frequency, and several breathing search processes comprising a first process and a second process. The first process takes as input a first sub-sample of a portion of the signal having a first duration. The second process takes as input a second sub-sample of a portion of the signal having a second duration. The first duration is greater than the second duration and the frequency of the first sub-sample is lower than the frequency of the second sub-sample.

[0006] Each process may include an autocorrelation analysis of the respective subsample to detect a period of repetition of a pattern and an associated reliability score.

[0007] Each subsample can include a respective set of measurements. The number of measurements in each subsample can be less than or equal to 180.

[0008] The first duration can be greater than or equal to 7 seconds and / or the second duration can be less than or equal to 7 seconds.

[0009] The breath search processes may include a third process. The third process may take as input a third subsample of a portion of the signal having a third duration. The second duration may be greater than the third duration and the frequency of the second subsample may be less than the frequency of the third subsample.

[0010] Each process can be repeated every X seconds with X being a real number less than or equal to 2 seconds.

[0011] The method may include generating an alert based on the results of the search processes.

[0012] A computer program for a vehicle UWB system is also provided. The computer program includes instructions for carrying out the method.

[0013] A UWB sensor of UWB system is also provided. The UWB sensor is configured to implement the method.

[0014] A vehicle UWB system is also provided. The UWB system comprises one or more such UWB sensors. The UWB system comprises a memory on which the computer program is stored. Brief description of the figures

[0015] Non-limiting examples will be described with reference to the following figures:

[0016] Illustrates a flowchart of an example of the process.

[0017] Illustrates examples of subsamples.

[0018] Illustrates an example of autocorrelation analysis.

[0019] Illustrates an example of a vehicle UWB system. Detailed description

[0020] A method implemented by a UWB system in a vehicle for detecting the breathing of one or more passengers is proposed. The UWB system comprises one or more UWB sensors. The method comprises, for each UWB sensor, an acquisition by the sensor of an echo-radar signal having an acquisition sampling frequency, and, several breathing search processes comprising a first process and a second process. The first process takes as input a first sub-sample of a portion of the signal having a first duration. The second process takes as input a second sub-sample of a portion of the signal having a second duration. The first duration is greater than the second duration and the frequency of the first sub-sample is lower than the frequency of the second sub-sample.

[0021] The process improves the detection of breathing of one or more passengers.

[0022] Indeed, the method allows the execution of a search process adapted to several types of passenger. In particular, the first process is adapted for a first type of passenger having a breathing frequency in a lower average frequency range than that of a second type of passenger, for which the second process is adapted. Each process can make it possible to detect a template of a (distinct) range of breathing frequencies of a (distinct) type of passenger. In particular, the frequency of the first subsample, which is lower than that of the second subsample, allows detection of the breathing frequency of the first type of passenger. The frequency of the second subsample, which is higher, allows detection of the breathing frequency of the second type of passenger.Each search process is therefore adapted for a particular type of passenger, which allows efficient detection of several types of passenger by the process.

[0023] In particular, each of the first and second search processes efficiently provides an accurate result. Indeed, since the first duration is greater than the second duration, and the frequency of the first subsample is lower than the frequency of the second subsample, the number of measurements of each subsample is optimized for each process. In particular, for each process, this number of measurements is optimized to guarantee a number of measurements allowing accurate detection while remaining reasonable to allow computational efficiency. Each of the first and second processes is therefore efficient and accurate.

[0024] Detecting the breathing of one or more passengers may be used to perform one or more vehicle functions. For example, the method may include activating one or more vehicle functions based on the results of the search processes. For example, the method may include closing the vehicle when none of the processes detects breathing in the vehicle, or not allowing the vehicle to close when the breathing of a passenger is detected in the vehicle by at least one process. For example, the method may include sending a signal initiating the activation of an alarm in the event that a breath corresponding to that of an infant is detected and is the only one while the vehicle is closed.For example, the method may include sending a signal initiating activation of an alarm in the event that breathing evolving into a state indicating potential driver drowsiness is detected while driving.

[0025] In examples, the method may include executing one or more safety processes based on the results of the search processes (detection or not and associated reliability score). For example, the method may include executing the one or more safety processes when the results indicate the presence of at least one passenger in the vehicle. For example, the method may include executing the one or more safety processes when at least one breathing search process detects the breathing of a passenger in the vehicle. For example, the method may include executing the one or more safety processes when a process searching for the breathing of children detects the breathing of a child, for example after the vehicle has been closed.Alternatively or additionally, the method may include executing the one or more safety processes based on a combination of the results of the search processes. For example, the method may include executing the one or more safety processes when the process searching for a child's breath detects the presence of a child, and a process searching for an adult's breath does not detect the presence of an adult, for example after the vehicle has been locked. This would mean that the child is alone in the vehicle, which represents a dangerous situation for the child.

[0026] An example of a safety process may include generating an alert. Generating the alert may include, for example, sending an alarm to the vehicle owner, for example, to their key or mobile phone. In other examples, generating the alert may include activating a siren or other audible signal emitted by the vehicle. Another example of a safety process may include maintaining a minimum and / or maximum temperature in the vehicle, for example, by controlling the vehicle's air conditioning and heating functions.

[0027] In other examples, the method may include, upon detecting a driver's breathing in the vehicle, activating one or more vehicle functions such as playing music or adjusting the mirrors appropriate for the driver. The results of the search processes may be used to perform any combination of these example functionality.

[0028] The method may include repeating the breath-search processes, for example during the execution of one or more vehicle functions. Each process may be repeated every X seconds, with X being a real number less than or equal to 2 seconds. The repetition of the breath-search processes allows for real-time detection of the breathing of the passenger(s). In addition, the repetition allows for rapid detection of a risky situation in order to quickly warn the owner of the vehicle. For each process, the portion of the signal on which the subsample is performed may last substantially until the next repetition of the process. For example, the portion may stop at the time of the repetition, or just before.The portion of the signal taken as input by each process can be a sliding time window that ends at the time of process execution, or just before (e.g. a time window of 5 to 10 seconds).

[0029] The method may repeat each of the processes after each elapse of the same predetermined duration for each of the processes. For example, the predetermined duration may be greater than or equal to the duration of the signal portion taken as input by the process which is the longest. In this case, each portion taken as input by each respiration search process may be, for each repetition, the last portion of the signal acquired at the time of the repetition. The method may comprise a repetition of each of the processes together (at the same time), or a repetition with a shift between the processes.

[0030] Alternatively, the method may repeat the processes independently of each other. In this case, the method may comprise, for each process, a repetition of the breath-searching process after the end of the previous repetition of the process, for example after the elapse of a respective predetermined duration greater than or equal to that of the portion of the signal taken as input by the process. The portions taken as input by each process at each repetition may follow one another.

[0031] For each UWB sensor, the method may execute the echo-radar signal acquisition and the breath-search processes concurrently. For example, the method may execute each of the breath-search processes during the echo-radar signal acquisition, after acquiring the signal portion taken as input by each of the breath-search processes. For each UWB sensor, the method may execute the breath-search processes concurrently, for example in parallel, or one after the other, successively.

[0032] The method may execute the echo-radar signal acquisition steps and the respiration search processes successively for each of the UWB sensors of the UWB system. For example, the UWB sensors may be numbered, and the method may execute the echo-radar signal acquisition steps and the respiration search processes for each of the UWB sensors by following the numbering of the UWB sensors, for example by executing these steps for a predetermined duration for each of the UWB sensors. Alternatively, the method may consider groups of UWB sensors, and the method may execute the echo-radar signal acquisition steps and the respiration search processes at the same time for the UWB sensors of the same group, and successively for each of the groups.Alternatively, the method may perform the echo radar signal acquisition steps and the respiration search processes for all UWB sensors of the UWB system at the same time.

[0033] The acquisition sampling frequency may be less than or equal to 250 Hz and / or greater than or equal to 20 Hz. The acquisition sampling frequency may correspond to the number of measurements per unit of time in the echo-radar signal (e.g., the number of measurements per second when the frequency is expressed in Hz). Similarly, the sub-sample frequency may correspond to the number of measurements of the sub-sample per unit of time (e.g., the number of measurements per second when the frequency is expressed in Hz). The acquisition sampling frequency may be greater than or equal to the sub-sample frequency of at least one respiration search process (e.g., greater than or equal to the sub-sample frequency of each of the processes).For example, the subsample frequency may be substantially equal to one-half of the acquisition sampling frequency, or one-third of the acquisition sampling frequency. The acquisition sampling frequency may be substantially equal to any multiple M of the subsample frequency, with M a real number. Alternatively or additionally, the subsample frequency of at least one breath search process may be equal to the acquisition sampling frequency. In this case, the subsample of the at least one breath search process may correspond to the sampling itself.

[0034] Each breath search process can search for a passenger type. Each passenger type can correspond to a category of human person (e.g., defined based on an age range) or to a category of animal (e.g., "dog" or "cat"). For example, the first process can search for the breath of an adult human, and the second process can search for the breath of a child human. For example, the third process can search for the breath of a dog, or of a child human in a lower age range than the second process. The subsample frequency of each process can be adapted according to the type of person the process is searching for. For example, the subsample frequency can correspond to a multiple of an average breathing frequency of the type of person the process is searching for.The subsample frequency can, for example, be equal to K times the average respiration rate, with K a real number greater than or equal to 2. Each respiration search process can detect a respiration having a frequency in a frequency range around an average respiration rate of the type of person the process is searching for. For example, a respiration search process searching for a child can detect a respiration having a frequency in a frequency range around 25 bpm (which corresponds to the average rate of a child). A respiration search process searching for an adult can detect a respiration having a frequency in a frequency range around 12 bpm.

[0035] Each process takes a subsample as input. Each process can perform the respiration search directly on the subsample. In other words, each process can exclude any pre-processing, for example can exclude (i.e., not execute) specific processing at a targeted frequency and / or amplitude range. For example, each process can exclude a micro-Doppler analysis. Each process can include an application of a function that takes as input all the measurements of the subsample. The function can be the same for each of the processes, or can be different for two or more processes. The function applied by each process can be constant for none of the measurements of the subsample, i.e., the function can be variable for each of the measurements of the subsample.

[0036] For example, each subsample can be a vector (x i) with i ranging from 1 to N. Each process can include the application of a function f. The result provided by the process for a subsample (x i ) can therefore bef (x i ), that is, the result of applying the function to the vector (x i ). The function f can be variable for each of the subsample measurements. In other words, whatever j is between 1 and N, there can be two different values ​​x j 1 and x j 2 for the x coordinate j and a value (x) j for the other coordinates (x i ) with i ranging from 1 to Neti ≠ j, for which f [x j 1, (x) j ] is different def [x j 2, (x) j ].

[0037] Each process may include an autocorrelation analysis of the respective subsample to detect a period of repetition of a pattern. The autocorrelation analysis may take as input the measurements of the respective subsample and determine as output whether a pattern is repeated in the signal portion from the measurements of the respective subsample. The repeated pattern may correspond to a breath of the passenger type that the process is looking for. The autocorrelation analysis may provide as output a period with which the pattern is repeated and / or a repetition frequency, which may correspond to the period and frequency of the detected breath.

[0038] The autocorrelation analysis can also output a reliability score associated with the detected pattern repeat period. The reliability score can quantify how reliably the pattern repeat is detected. The reliability score can indicate how accurate the breath detection is. For example, the reliability score can be a real number between 0 and 1, with a value of 0 representing low or no reliability in detection, and 1 representing absolute reliability in detection. The autocorrelation analysis of the respective subsample can detect a pattern repeat period and an associated reliability score using the method described in Section 2.2.2 of “On Periodicity Detection and Structural Periodic Similarity,” Michail Vlachos, Philip Yu, and Vittorio Castelli, IBM TJ Watson Research Center (URL: http: / alumni.cs.ucr.edu / ~mvlachos / pubs / sdm05.pdf), which is incorporated herein by reference.

[0039] For example, the first subsample can be a vector (x i )1 and the second subsample a vector (x i )2. The number of coordinates of each of the vectors (x i )1and (x i )2 can be the same. The coordinates of the vectors (x i )1and (x i )2 can represent the measurements of the first subsample and the second subsample respectively. For example, each vector can include two coordinates for each measurement, the first coordinate indicating the measurement time and the second an amplitude of the signal at the measurement time. The coordinates of the vectors (x i )1 can represent measurements for which the measurement time is during the first duration and the coordinates of the vector (x i )2 of the measurements for which the measurement time is during the second duration. The measurements represented by each of the vectors (x i )1and (x i)2 can for example be regularly spaced over the first duration and the second duration respectively.

[0040] Autocorrelation analysis may include applying an autocorrel[] function. The first process may include applying the autocorrel[] function to the first vector (x i )1and the result of the first process can therefore be autocorrel[(x i )1]. The second process may include applying the autocorrel[] function to the second vector (x i )2and the result of the first process can therefore be autocorrel[(x i)2]. The autocorrel[] function may be a statistical function. The autocorrel[] function may be the autocorrelation function described in Section 2.2.2 of “On Periodicity Detection and Structural Periodic Similarity,” Michail Vlachos, Philip Yu, and Vittorio Castelli, IBM TJ Watson Research Center (URL: http: / alumni.cs.ucr.edu / ~mvlachos / pubs / sdm05.pdf), which is incorporated herein by reference.

[0041] The subsample of each process may comprise a respective set of measurements. The number of measurements in the set may be a function of the subsample frequency and the duration presented by the portion of the signal taken as input by the respiration search process. The number of measurements may be equal to the result of multiplying the subsample frequency by the duration presented by the portion of the signal. The numbers of measurements of the subsamples may be within the same interval. For example, the number of measurements in each subsample may be less than or equal to 180. The number of measurements in each subsample may be greater than or equal to 65.

[0042] In examples, the breath search processes include a third process. The third process takes as input a third subsample of a portion of the signal having a third duration. The second duration is greater than the third duration and the frequency of the second subsample is less than the frequency of the third subsample. In still other examples, the breath search processes may form a set of N breath search processes with N a real number. Each process in the set may take as input a subsample SE i of a portion of the signal having a duration i and with a frequencyf i . The duration i and the frequencyf i of each processi, starting from the second process (i > 1), can be so thatf i >f i -1 etd i < d i -1 , withf i -1 etd i-1 respectively the frequency and duration of the subsampleSE i-1 of processi -1 (the frequencyf1 and the durationd1 of the first process being predetermined). The number of processes executed for each of the UWB sensors can be the same for each UWB sensor. Alternatively, the number of processes executed can be different for at least two UWB sensors.

[0043] In examples, the frequency of the first subsample may be between 1 and 12 Hz. The first duration may be greater than or equal to 7 seconds. The first duration may be less than or equal to 10 seconds. The frequency of the first subsample and the first duration may be between these bounds in each of the two implementation examples given below. The first process may detect the breathing of an adult human.

[0044] In a first example implementation, the breath search processes may consist of two breath search processes (a first and a second process). In this first example implementation, the frequency of the second subsample may be between 4 and 110 Hz. The second duration may be less than or equal to 7 seconds. The second duration may be greater than or equal to 6 seconds. In this first example, the first process may detect the breathing of an adult human, and the second process the breathing of a child.

[0045] In a second example implementation, the breath search processes may consist of three breath search processes (a first process, a second process, and a third process). In this second example implementation, the frequency of the second subsample may be between 4 and 50 Hz. The second duration may be less than or equal to 7 seconds. The second duration may be greater than or equal to 6 seconds. The frequency of the third subsample may be between 15 and 110 Hz, and the third duration may be between 4 and 7 seconds. In this second example, the first process may detect the respiration of an adult human, the second process the respiration of an adolescent child, and the third process the respiration of an infant.

[0046] Each UWB sensor may use the UWB communication protocol, for example that specified by IEEE 802.15.4. Each UWB sensor may be configured to perform echo-radar signal acquisition and respiration search processes. For example, each UWB sensor may comprise means for transmitting and receiving the echo-radar signal, a processor and a memory (for example non-volatile). Instructions for performing the echo-radar signal acquisition and respiration search processes may be recorded on the memory. The method may record the acquired echo-radar signal on the memory. The processor may perform the echo-radar signal acquisition and respiration search processes by executing the instructions recorded on the memory and from the echo-radar signal recorded, after acquisition, also on the memory.

[0047] A computer program for a vehicle UWB system is also provided. The computer program may include instructions for performing the method. For example, the computer program may include instructions for each of the UWB sensors in the system. The instructions may be for performing the method when said program is executed by a UWB sensor processor. The instructions may be stored in non-volatile memory (e.g., the memory of the UWB sensor).

[0048] A vehicle UWB system is also provided. The UWB system comprises a memory on which the computer program is recorded. The UWB system may comprise a processor for executing said program. The UWB system may be configured to implement the method. The UWB system may comprise one or more UWB sensors. The UWB system may also comprise a central computer. The central computer may comprise the memory and the processor of the UWB system. The UWB system may also comprise connection means between the central computer and / or each of the one or more UWB sensors (e.g. cables connecting the central computer with each of the UWB sensors).

[0049] Examples will now be given with reference to Figures 1 to 3.

[0050] Illustrates a flowchart of an example of the method. The method is implemented by a UWB system in a vehicle for detecting the breathing of one or more passengers. The UWB system comprises one or more UWB sensors. The method comprises, for each UWB sensor, an acquisition S10 by the sensor of an echo-radar signal having an acquisition sampling frequency, and several breathing search processes S20. The breathing search processes S20 consist of a first process S21, a second process S22 and a third process S23. The first process S21 takes as input a first subsample of a portion of the signal having a first duration. The second process S22 takes as input a second subsample of a portion of the signal having a second duration. The third process S23 takes as input a third subsample of a portion of the signal having a third duration.

[0051] The first process S21 uses a large window, i.e., the first duration is greater than the second duration and the third duration. The second process S22 uses a medium window, i.e., the second duration is less than the first duration and greater than the third duration. The third process S23 uses a small window, i.e., the third duration is less than the first duration and the second duration. The first duration is greater than the second duration, and the frequency of the first subsample is less than the frequency of the second subsample. The second duration is greater than the third duration, and the frequency of the second subsample is less than the frequency of the third subsample.

[0052] The first process S21 includes a low respiration frequency search module. The first process S21 therefore searches for a passenger type having a low average frequency. For example, the first process S21 can detect the respiration of an adult passenger. For example, the first process S21 can detect the respiration of a passenger having a respiration rate less than or equal to 20 bpm (beats per minute), for example around 12 bpm. The first process S21 is based on a first subsample with a low frequency. The first process S21 includes an autocorrelation analysis of the first subsample to detect a period of repetition of a pattern and determine an associated reliability score.

[0053] The second process S22 includes a mean respiration frequency search module. The second process S22 therefore searches for a passenger type having a higher mean frequency than the first process S21. For example, the second process S22 can detect the respiration of a teenage child. For example, the second process S22 can detect the respiration of a passenger having a respiration frequency higher than 20 bpm and / or lower than 25 bpm, for example around 22 bpm. The second process S22 is based on a second subsample with an intermediate frequency. The second process S22 includes an autocorrelation analysis of the second subsample to detect a period of repetition of a pattern and determine an associated reliability score.

[0054] The third process S23 includes a high respiration frequency search module. The third process S23 therefore searches for a passenger type with a higher average frequency than the first process S21 and the second process S22. For example, the third process S23 can detect the respiration of a young child (e.g., an infant). For example, the third process S23 can detect the respiration of a passenger with a respiration frequency higher than 25 bpm, for example, around 27 bpm. The third process S23 is based on a third subsample with a high frequency. The third process S23 includes an autocorrelation analysis of the third subsample to detect a period of repetition of a pattern and determine an associated reliability score.

[0055] The method then comprises executing a decision layer based on the results of the processes. For example, the decision layer may identify a risky situation based on the results of the processes. The execution of the decision layer may, for example, identify the presence of a child alone in the vehicle. For this, the method may determine that a child is alone in the vehicle when the first process, detecting adult passengers, does not detect the presence of an adult passenger in the vehicle (for example with an associated reliability score greater than a predetermined threshold), and at least one of the second and third processes, detecting the breathing of adolescent child or infant passengers, detects the presence of at least one adolescent child or infant passenger (for example also with an associated reliability score greater than a predetermined threshold).

[0056] The decision layer may also take vehicle information as input. For example, the decision layer may take a vehicle lock status as input. The decision layer may condition the identification of the risk situation based on this vehicle information. For example, the decision layer may identify a risk situation when it is determined that a child is alone in the vehicle and the vehicle is locked. When a risk situation is identified, the method may include performing one or more safety processes, such as generating an alert.

[0057] Laillustrates examples of subsamples. Lashows a first example of a subsample 100 of a portion of the signal 101. This first subsample 100 has a first duration 102. This first subsample 100 can for example be used by the first process S21 of the. The first subsample comprises a set of measurements 103. The first duration 102 represents approximately two periods of the detected passenger's breathing. The number of measurements of the first subsample 100 allows detection of the breathing of an adult passenger by the first process. In particular, the number of measurements to be analyzed by the first process is not excessive, in particular given the length of the first duration 102. The frequency of the first subsample is adapted to have a reasonable number of measurements to be analyzed by the first process (while having several measurements per breathing period).This ensures the accuracy and efficiency of the first process to detect the breathing of an adult passenger.

[0058] La also shows a second example of a subsample 200 of a portion of the signal 201. This second subsample 200 has a second duration 202. This second subsample 200 can for example be used by the second process S22 of the. The second subsample comprises a set of measurements 203. The second duration 202 represents approximately two periods of the detected passenger's breathing. The number of measurements of the second subsample 200 allows detection of the breathing of a child passenger by the second process. In particular, the second subsample 200 comprises several measurements, although the second duration 202 is shorter than the first duration 102, and the number of measurements allows detection of the child's breathing.The frequency of the second subsample is higher than that of the first process so that the second subsample 200 includes multiple measurements per breathing period of a child passenger (which is shorter than that of an adult passenger). This ensures the accuracy and efficiency of the second process for detecting the breathing of a child passenger.

[0059] illustrates an example of the autocorrelation analysis. In this example, the echo-radar signal to be analyzed comprises two time components: the amplitude and the phase, which are represented on the by the curves 401 and 402 respectively. shows extracts of the two time components of echo-radar signals in two breathing scenarios 410 and 420. The autocorrelation analysis of the respective subsample can detect a period of repetition of a pattern. The breathing patterns observed on the amplitude component generally share an underlying characteristic, namely a shape close to a triangular signal whose high and low peaks are flattened so as to form two plateaus (which may be of different durations), and whose rising slope is steeper (and therefore shorter in time) than the falling slope, which may take a slight convex aspect. The autocorrelation analysis can include a detection of a repetition of these patterns.The curves in this figure are from a recording of respirations under optimal conditions for detection. In other examples, the raw echo-radar signal may contain more noise and the method may include filtering the phase and amplitude of the echo-radar signal (the autocorrelation analysis may then use the filtered signal). For example, the method may include applying a bandpass filter by adjusting the cutoff frequencies according to the range of respiration frequencies sought.

[0060] Each process may include a periodicity estimate from the application of the autocorrelation function to the amplitude or to a complex representation of the signal. The autocorrelation function may be the AutoCorrelation Function described in Section 2.2.2 of “On Periodicity Detection and Structural Periodic Similarity,” Michail Vlachos, Philip Yu, and Vittorio Castelli, IBM TJ Watson Research Center (URL: http: / alumni.cs.ucr.edu / ~mvlachos / pubs / sdm05.pdf), which is incorporated herein by reference. Each process may provide the value of the autocorrelation function directly as a reliability score for the corresponding period.

[0061] In other examples, the method may include aggregating information from multiple sensors (e.g., using clustering methods). The aggregation of information makes it possible to robustify the detection and to mitigate the risk of multiples of the detected period appearing in the estimation via the autocorrelation function.

[0062] Illustrates an example of a vehicle UWB system 30. The UWB system 30 comprises a network comprising UWB sensors 31 and a centralized electronic control unit 33. The UWB sensors 31 are positioned at the front of the vehicle, at the rear of the vehicle or inside the vehicle (for example in the cabin). The network also comprises connection means 32 (for example cables) between the UWB sensors 31 and the centralized electronic control unit 33.

[0063] The UWB system 30 is configured to implement the method. For this, each UWB sensor 31 may be configured to perform the acquisition of the echo-radar signal and the breathing search processes. For example, each UWB sensor 31 may comprise means for transmitting and receiving echo-radar signals, a processor and a memory (for example non-volatile) on which each echo-radar signal received and instructions for executing the processes may be recorded. Each UWB sensor 31 may be configured to send the results of the processes it has executed to the centralized electronic control unit 33, via the connection means 32.The processor of the centralized electronic control unit 33 can then, from the results received from each of the UWB sensors, execute the decision-making layer to identify a risky situation (for example that a child is alone in the vehicle while the vehicle is closed), and when such a situation is identified, generate an alert.

Claims

Method implemented by a UWB system (30) in a vehicle for detecting breathing of one or more passengers, the UWB system (30) comprising one or more UWB sensors (31), the method comprising, for each UWB sensor (31): an acquisition (S10) by the sensor of an echo-radar signal having an acquisition sampling frequency; andmultiple breath search processes (S20) comprising a first process (S21) and a second process (S22), the first process (S21) taking as input a first subsample (100) of a portion of the signal (101) having a first duration (102), the second process (S22) taking as input a second subsample (200) of a portion of the signal (201) having a second duration (202), the first duration (102) being greater than the second duration (202) and the frequency of the first subsample (100) being lower than the frequency of the second subsample (200). The method of claim 1, wherein each process comprises an autocorrelation analysis of the respective subsample to detect a period of repetition of a pattern and an associated reliability score. The method of claim 1 or 2, wherein each subsample comprises a respective set of measurements (103, 203), the number of measurements (103, 203) in each subsample being less than or equal to 180. Method according to any one of claims 1 to 3, wherein the first duration (102) is greater than or equal to 7 seconds and / or the second duration (202) is less than or equal to 7 seconds. A method according to any one of claims 1 to 4, wherein the breath search processes comprise a third process (S23), the third process (S23) taking as input a third subsample of a portion of the signal having a third duration, the second duration (202) being greater than the third duration and the frequency of the second subsample (200) being less than the frequency of the third subsample. A method according to any one of claims 1 to 5, wherein each process is repeated every X seconds with X being a real number less than or equal to 2 seconds. A method according to any one of claims 1 to 6, the method comprising generating (S40) an alert based on the results of the search processes. A computer program for a vehicle UWB system (30) comprising instructions for carrying out the method of any one of claims 1 to 7. UWB sensor (31) of UWB system (30) configured to implement the method according to any one of claims 1 to 7. UWB system (30) comprising one or more UWB sensors (31) according to claim 9, the UWB system (30) comprising a memory on which the program according to claim 8 is recorded.

Images