[go: up one dir, main page]

WO2024112265A1 - Traitement de données chronologiques et systèmes et procédés associés - Google Patents

Traitement de données chronologiques et systèmes et procédés associés Download PDF

Info

Publication number
WO2024112265A1
WO2024112265A1 PCT/SG2023/050782 SG2023050782W WO2024112265A1 WO 2024112265 A1 WO2024112265 A1 WO 2024112265A1 SG 2023050782 W SG2023050782 W SG 2023050782W WO 2024112265 A1 WO2024112265 A1 WO 2024112265A1
Authority
WO
WIPO (PCT)
Prior art keywords
time
data
series data
processed
series
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Ceased
Application number
PCT/SG2023/050782
Other languages
English (en)
Inventor
Haihong Zhang
Chuanchu Wang
Kai Keng Ang
Soon Huat Ng
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Agency for Science Technology and Research Singapore
Original Assignee
Agency for Science Technology and Research Singapore
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Agency for Science Technology and Research Singapore filed Critical Agency for Science Technology and Research Singapore
Publication of WO2024112265A1 publication Critical patent/WO2024112265A1/fr
Anticipated expiration legal-status Critical
Ceased legal-status Critical Current

Links

Classifications

    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04Q—SELECTING
    • H04Q9/00—Arrangements in telecontrol or telemetry systems for selectively calling a substation from a main station, in which substation desired apparatus is selected for applying a control signal thereto or for obtaining measured values therefrom
    • H04Q9/04—Arrangements for synchronous operation
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00—Measuring for diagnostic purposes; Identification of persons
    • A61B5/02—Detecting, measuring or recording for evaluating the cardiovascular system, e.g. pulse, heart rate, blood pressure or blood flow
    • A61B5/026—Measuring blood flow
    • A61B5/0261—Measuring blood flow using optical means, e.g. infrared light
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00—Measuring for diagnostic purposes; Identification of persons
    • A61B5/24—Detecting, measuring or recording bioelectric or biomagnetic signals of the body or parts thereof
    • A61B5/316—Modalities, i.e. specific diagnostic methods
    • A61B5/369—Electroencephalography [EEG]
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00—Measuring for diagnostic purposes; Identification of persons
    • A61B5/72—Signal processing specially adapted for physiological signals or for diagnostic purposes
    • A61B5/7203—Signal processing specially adapted for physiological signals or for diagnostic purposes for noise prevention, reduction or removal
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00—Measuring for diagnostic purposes; Identification of persons
    • A61B5/72—Signal processing specially adapted for physiological signals or for diagnostic purposes
    • A61B5/7235—Details of waveform analysis
    • A61B5/7246—Details of waveform analysis using correlation, e.g. template matching or determination of similarity
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00—Measuring for diagnostic purposes; Identification of persons
    • A61B5/72—Signal processing specially adapted for physiological signals or for diagnostic purposes
    • A61B5/7271—Specific aspects of physiological measurement analysis
    • A61B5/7285—Specific aspects of physiological measurement analysis for synchronizing or triggering a physiological measurement or image acquisition with a physiological event or waveform, e.g. an ECG signal
    • G—PHYSICS
    • G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
    • G16H—HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
    • G16H40/00—ICT specially adapted for the management or administration of healthcare resources or facilities; ICT specially adapted for the management or operation of medical equipment or devices
    • G16H40/60—ICT specially adapted for the management or administration of healthcare resources or facilities; ICT specially adapted for the management or operation of medical equipment or devices for the operation of medical equipment or devices
    • G16H40/67—ICT specially adapted for the management or administration of healthcare resources or facilities; ICT specially adapted for the management or operation of medical equipment or devices for the operation of medical equipment or devices for remote operation
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00—Measuring for diagnostic purposes; Identification of persons
    • A61B5/0002—Remote monitoring of patients using telemetry, e.g. transmission of vital signals via a communication network
    • A61B5/0004—Remote monitoring of patients using telemetry, e.g. transmission of vital signals via a communication network characterised by the type of physiological signal transmitted
    • A61B5/0006—ECG or EEG signals
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00—Measuring for diagnostic purposes; Identification of persons
    • A61B5/0002—Remote monitoring of patients using telemetry, e.g. transmission of vital signals via a communication network
    • A61B5/0015—Remote monitoring of patients using telemetry, e.g. transmission of vital signals via a communication network characterised by features of the telemetry system
    • A61B5/0022—Monitoring a patient using a global network, e.g. telephone networks, internet
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04Q—SELECTING
    • H04Q2209/00—Arrangements in telecontrol or telemetry systems
    • H04Q2209/80—Arrangements in the sub-station, i.e. sensing device
    • H04Q2209/84—Measuring functions
    • H04Q2209/845—Measuring functions where the measuring is synchronized between sensing devices

Definitions

  • the present invention relates, in general terms, to a system for processing time-series data such as electroencephalogram (EEG) data and photoplethysmogram (PPG) data.
  • EEG electroencephalogram
  • PPG photoplethysmogram
  • Disclosed herein is a method to synchronize data from multiple sources. It is suitable for both offline data analysis and model training and for online data processing applications.
  • the present invention provides a system for producing synchronised time-series data, the system comprising: one or more processor(s), a memory comprising program code executable by the processor(s): receive a first time-series data from a first measurement device and a second time-series data from a second measurement device, wherein the first and second time-series data comprise a series of data points and an arrival timestamp for each data point; process the received first and second time-series data using a regression model to obtain processed first and second time-series data respectively, wherein the regression model performs regression on the received arrival timestamps of the first and second time-series data to generate the processed first and second time-series data comprising synchronised timestamps, thereby synchronising the processed first and second timeseries data with respect to each other.
  • a method for producing synchronised time-series data comprising: receiving a first time-series data from a first measurement device and a second time-series data from a second measurement device, wherein the first and second time-series data comprise a series of data points and associated arrival timestamps; processing the received first and second time-series data using a regression model to obtain processed first and second time-series data respectively, wherein the regression model performs regression on the received arrival timestamps of the first and second time-series data to generate the processed first and second time-series data comprising synchronised timestamps, thereby synchronising the processed first and second time-series data with respect to each other.
  • Figure 1 illustrates a method for producing synchronised timeseries data, along with a linear regression algorithm implemented in that method, in accordance with present teachings
  • Figure 2 shows an example of input, received and linear- regression-based corrected data points
  • Figures 3(a) to 3(d) show a delay simulation of two sine wave data in which Figure 3(a) is the simulated delay generated by a random model, with mean delay of 1.5 seconds, blue for signal 1 and red for signal 2, image Figure 3 (b) shows the simulated received data with generated delay values, image Figure 3 (c) shows received data processed with our linear regression algorithm, and image Figure 3 (d) plots the difference between the two signals, blue for received data with S2 interpolate with SI timestamps, without any processing, red for Linear Regression processed data, with S2 interpolated with SI timestamps;
  • Figure 4 shows the mean squared error (MSE) with randomize index changes from 0.1 to 1 and system running over 100 cycles;
  • Figure 5 shows data points generated by a Bluetooth device measuring both EEG and PPG with different sampling rates
  • Figure 6 shows the same data points as those shown in Figure 5, but after linear regression-based synchronisation in accordance with present teachings.
  • Figure 7 shows a system for implementing the method of Figure 1.
  • Data synchronisation techniques used herein employ regression techniques for merging timestamps for data from multiple sources. In general, these techniques leverage linear regression. Typically, linear regression algorithms are designed for one source of data. The present teachings extend linear regression to multiple data source synchronization.
  • Figure 1 illustrates a method 100 for producing synchronised timeseries data. Synchronisation is based on the understanding that, in general, a measurement device will generate samples at a fixed sampling rate. The term “samples” will be interchangeably used with “data points”.
  • the method 100 involves receiving the time series data (step 102) and processing it to obtain processed time series data (step 104). Given the method 100 synchronises data across multiple devices, step 102 involves receiving first and second time-series data from first and second measurement devices, respectively.
  • the same teachings can be extended to three or more devices either by synchronising using the timestamps of one of the multiple devices, or synchronising using the system clock.
  • the first and second time-series data comprise a series of data points and an arrival timestamp for each data point.
  • Step 104 then involves processing the received first and second time-series data using a data processing model.
  • This output is processed first and second time-series data.
  • the timestamps of one time series data for illustration purposes, the first time series data
  • the first time series data may be the same as the processed first time series data - i.e. it remains unchanged.
  • the data processing model is a regression model that performs regression on the received arrival timestamps of the first and second time-series data.
  • the regression model generates the processed first and second time-series data such that they have synchronised timestamps. As a result, the regression model synchronises the processed first and second time-series data with respect to each other.
  • System paced synchronisation uses a fixed system time step, e.g. generated by the system clock, to output data samples. There is no master data source, and all data sources may need to have timestamps interpolated. Advantageously, however, the output-sampling rate is fixed.
  • Each form of synchronisation takes the same data inputs, and the data are asynchronously supplied, since every data source has its own sampling rate and delays on the time that data takes to be received by the system running method 100.
  • Step 104 may comprise passing each time-series data to a corresponding, or separate, linear regression instance for producing the processed timestamps.
  • each linear regression instance runs on a processor associated with memory.
  • the memory stores a queue of the time-series data for later synchronisation. This can be used to match the data with the processed (calculated) timestamps and/or to maintain a sliding window of data that is buffered for synchronisation with data points that are subsequently received.
  • the output data is set based on the master data source.
  • a number of samples are acquired for one data source (the master data source), the total number of samples is determined and corresponding timestamps are calculated. This can be done by interpolating over a known time window. For example, if 60 data points are received in a one second window, a 60Hz sampling rate will be assumed even if some samples are yet to be received, and the sequence of data points are allocated timestamps of successive intervals of l/60 th of a second.
  • next-ts is set to store the timestamp of the next output sample: next-ts.
  • NAN not a number
  • the algorithm 106 receives an input stream of time-series data y. y is then stored in a buffer. A count c is taken of the number of data points in y. If c is less than a current length of data used for regression, xlen, (i.e. current number of data points, such as the number of data points from the master data source in the list), then regression is performed over the count c for the time period covered by xlen. In this case De is used to evaluate if there is sufficient data for accurate regression model computation. Then timestamp for the first time entry for the data source yl is set as the timestamp used for obtaining the single data point. If the count is greater than xlen but less than the length of the buffer (set empirically), then the time step Xstep for the subsequent sample will be set to be c/xlen.
  • Ysum is given by: where j is the most recent sample in input stream y at timestep and N is the total number of timesteps over which regression takes place.
  • Linear regression process is applied to data arrival time - i.e. the time at which a data point is received at the system implementing method 100. There is an assumption that at the device site, the sampling rate is fixed. The variances in arrival time are caused by data processing and transmission time. [0033] The system will thus receive data point x at time y as below:
  • a linear-regression function takes data sample arrival timestamp (the time at which the data was received) as input, and returns the processed timestamp.
  • a data buffer is created to keep an historical record of a sliding window of data points for parameter updating.
  • step-size parameter, xstep the number of data points per regression parameter update - i.e. the number of data points before the regression model is recomputed, bearing in mind that regression over a continuous stream of data cannot be done in real time over the entire data set and the present methodology makes accurate regression by windowing the data stream every xstep data points).
  • step-size parameter, xstep the number of data points per regression parameter update - i.e. the number of data points before the regression model is recomputed, bearing in mind that regression over a continuous stream of data cannot be done in real time over the entire data set and the present methodology makes accurate regression by windowing the data stream every xstep data points.
  • xstep the number of data points per regression parameter update - i.e. the number of data points before the regression model is recomputed, bearing in mind that regression over a continuous stream of data cannot be done in real time over the entire data set and the present methodology makes accurate regression by windowing the data stream every xstep data points.
  • the sampling rate is 250 Hz.
  • the original data is shown in thick solid line, which is a standard 2 Hz sine signal. Due to the transmission and buffering latency, the end side received data in thin solid line shows zigzag shapes when plotted against the time (axis X). Applying linear regression algorithm to the received timestamps, the processed data shown in dashed line has shape very close the original data, except a few data at the beginning not shown the smooth signal, due to too few data points contributing to the regression at beginning. After a very short while (around 100 milliseconds), the output becomes stable, and the processed data (orange) shows very similar shape to the original data, with a short delay in time.
  • T1 and data XI are generated for signal 1
  • time sequence T2 and data X2 are generated for signal 2.
  • the differences are calculated between XI and interpolated X2 at SI time sequences (i.e. the timestamps corresponding to master data source, Tl).
  • SI time sequences i.e. the timestamps corresponding to master data source, Tl.
  • the solid line curve shows the differences of XI and X2 with Trcv_2 interpolated at Trcv_l, and plotted against time sequence Trcv_l
  • the dashed line shows the differences between XI and X2 with tr2 interpolated at tri, plotted against time sequence tri.
  • MSE of XI and interpolated X2 are calculated to evaluate the linear-regression algorithm.
  • Figure 4 shows the effect, over 100 cycles of operation, of changing the delay generation model randomization parameter from 0.1 to 1, with the MSE calculated for individual rates.
  • the solid line shows the mean and standard derivation of the MSE for received data without linear regression and the dashed line shows the mean of stand derivation of the MSE for data points processed using the linear regression algorithm.
  • the error deteriorates (increases) as randomisation increases, the output nevertheless tracks the original, predelay signal, far more closely than in the absence of regression analysis.
  • the data averages 0.5 error which is understandable as it reflects that average between 0 and the maximum error of 1.
  • the lower accuracy rate may be due to data points arriving so early, or late, when compared with neighbouring data points, that they are included in a time window that is earlier, or later, than the time window in which they would otherwise have been processed.
  • the method 100 operates on electroencephalogram (EEG) and photoplethysmogram (PPG) data.
  • EEG electroencephalogram
  • PPG photoplethysmogram
  • a wireless EEG device captures a forehead EEG signal as well as PPG signal, sampling at different rates: 256Hz and 60Hz.
  • PPG and EEG there are also accelerometer and gyroscope sensor data that can be captured, as well as temperature, battery level and other kinds of data. All data are sensed at different sampling rates, as reflected in Figure 5. All data are transferred to the system (e.g. that shown in Figure 7) through Bluetooth LE wireless communication protocols.
  • the EEG and PPG data and battery level of the measuring device are captured.
  • the EEG and PPG data facilitate real time display and processing.
  • Figure 5 shows the data points prior to regression being applied. It is obvious that there are gaps in the data being obtained in the PPG data channel 500, between consecutive packages. This makes the displayed PPG signal zig-zagged. Consequently, the PPG data cannot be used directly for analysis and application.
  • FIG. 7 shows a system implementing the method 100.
  • the system 700 comprises at least one processor 702, memory 704 and a network interface 706 for communicating with measurement devices 708 over a wired or wireless network 710.
  • the memory comprises program code or instructions that, when executed by processor(s) 702 implement the method 100.
  • the program code 712 implements a time-series data processing module 714 that runs a regression model 716 to process received time series data. There may be a separate regression model for each time-series data input.
  • the system 700 may be any suitable device, such as a personal computer, smartphone or other device.
  • the separate program code modules and models may be implemented in any desired manner, such as in hardware or software.
  • the system 700 has been shown as a single device, it, or components thereof, may be distributed across multiple servers.

Landscapes

  • Health & Medical Sciences (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Engineering & Computer Science (AREA)
  • Biomedical Technology (AREA)
  • Public Health (AREA)
  • General Health & Medical Sciences (AREA)
  • Medical Informatics (AREA)
  • Molecular Biology (AREA)
  • Surgery (AREA)
  • Biophysics (AREA)
  • Pathology (AREA)
  • Veterinary Medicine (AREA)
  • Heart & Thoracic Surgery (AREA)
  • Physics & Mathematics (AREA)
  • Animal Behavior & Ethology (AREA)
  • Signal Processing (AREA)
  • Physiology (AREA)
  • Psychiatry (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Artificial Intelligence (AREA)
  • Epidemiology (AREA)
  • General Business, Economics & Management (AREA)
  • Business, Economics & Management (AREA)
  • Primary Health Care (AREA)
  • Hematology (AREA)
  • Cardiology (AREA)
  • Psychology (AREA)
  • Computer Networks & Wireless Communication (AREA)
  • Complex Calculations (AREA)

Abstract

La divulgation concerne un système et un procédé de production de données chronologiques synchronisées. Le système reçoit des premières données chronologiques provenant d'un premier dispositif de mesure et des secondes données chronologiques provenant d'un second dispositif de mesure. Chaque donnée chronologique comprend une série de points de données et une estampille temporelle d'arrivée pour chaque point de données. Le système traite les données chronologiques au moyen d'un modèle de régression pour obtenir des données chronologiques traitées. Le modèle de régression met en oeuvre une régression sur les estampilles temporelles d'arrivée reçues des données chronologiques pour générer les données chronologiques traitées, qui comprennent ensuite des estampilles temporelles synchronisées. De cette manière, les premières et secondes données chronologiques traitées sont synchronisées les unes par rapport aux autres.
PCT/SG2023/050782 2022-11-25 2023-11-24 Traitement de données chronologiques et systèmes et procédés associés Ceased WO2024112265A1 (fr)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
SG10202260209Q 2022-11-25
SG10202260209Q 2022-11-25

Publications (1)

Publication Number Publication Date
WO2024112265A1 true WO2024112265A1 (fr) 2024-05-30

Family

ID=91196711

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/SG2023/050782 Ceased WO2024112265A1 (fr) 2022-11-25 2023-11-24 Traitement de données chronologiques et systèmes et procédés associés

Country Status (1)

Country Link
WO (1) WO2024112265A1 (fr)

Citations (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112636859A (zh) * 2020-12-23 2021-04-09 西安云维智联科技有限公司 基于线性回归算法的ieee1588协议时间校准方法
EP3865059A1 (fr) * 2020-02-13 2021-08-18 Qompium Procédé mis en uvre par ordinateur pour la synchronisation d'un signal de photopléthysmographie (ppg) à l'aide d'un signal d'électrocardiogramme (ecg)

Patent Citations (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
EP3865059A1 (fr) * 2020-02-13 2021-08-18 Qompium Procédé mis en uvre par ordinateur pour la synchronisation d'un signal de photopléthysmographie (ppg) à l'aide d'un signal d'électrocardiogramme (ecg)
CN112636859A (zh) * 2020-12-23 2021-04-09 西安云维智联科技有限公司 基于线性回归算法的ieee1588协议时间校准方法

Similar Documents

Publication Publication Date Title
US11917567B2 (en) Method and device for synchronizing data packets from embedded data sensors monitoring body motion of a patient
US7379480B2 (en) Fast frequency adjustment method for synchronizing network clocks
CN106253902B (zh) 具有多器件同步复位识别校正功能的多通道并行采集系统
US11757614B2 (en) Accurate timestamp correction
US9374214B2 (en) Communication apparatus, communication system, and communication method
JP4032421B2 (ja) 測定データ同期システム
CN104243129B (zh) 接收器、计算时间差的方法以及计算机可读存储介质
CN103141039A (zh) 用于在通讯网络中的时间同步的方法
WO2016168063A1 (fr) Procédés, systèmes et supports lisibles par ordinateur pour synchroniser une temporisation entre des cartes d'interface réseau (nic) dans un dispositif de test d'équipement de réseau
JP6827296B2 (ja) データ通信方法
WO2024112265A1 (fr) Traitement de données chronologiques et systèmes et procédés associés
CN105303012A (zh) 一种基于区间灰数的异步等数据率航迹关联算法
CN107800529A (zh) 一种网络节点的时钟频率同步方法
CN113049870B (zh) 消除触发抖动的触发信号处理方法及触发信号处理装置
JP6683137B2 (ja) 産業用コントローラシステム
JP2003242130A (ja) 収集データの同期化方法及びデータ処理システム
US8775852B2 (en) Method for sensing input signal changes
KR100894358B1 (ko) 단말기의 시각 동기화 방법
JP6953735B2 (ja) 情報処理装置、情報処理システム、情報処理方法及び情報処理プログラム
US12314079B2 (en) Synthetic timestamp compensating for skew between system clock and sensor clock
JP2009200787A (ja) クロック周波数安定化方法およびデータ受信装置
CN112993949A (zh) 一种用于输电线路纵联差动保护采样的同步方法及装置
CN119108101A (zh) 一种生理数据处理方法、装置、电子设备及存储介质
JP6823701B1 (ja) クロック調整装置、方法、及びプログラム
JP4698107B2 (ja) 改良型ベイズ・カルマン時間推定装置及び方法

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 23895145

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

WWE Wipo information: entry into national phase

Ref document number: 11202503385T

Country of ref document: SG

WWP Wipo information: published in national office

Ref document number: 11202503385T

Country of ref document: SG

122 Ep: pct application non-entry in european phase

Ref document number: 23895145

Country of ref document: EP

Kind code of ref document: A1