The 4th International Conference on Awareness Science and Technology August 21-24, 2012, Korea University, Seoul, Korea Towards Smart Health Monitoring System for Elderly People Achraf Ben Ahmed, Yumiko Kimezawa, Abderazek Ben Abdallah Graduate School of Computer Science and Engineering, Adaptive Systems Laboratory The University of Aizu, Aizu-Wakamatsu ,Japan Email:m5151161@u-aizu.ac.jp 9/27/2012 1
Contents • Background • Contributions – Period-Peak Detection (PPD) Algorithm – System architecture System architecture The interactive real-time interface (IRI) • Design and evaluation results •Conclusion and future work 9/27/2012 2
Contents • Background • Contributions – Period-Peak Detection (PPD) Algorithm – System architecture System architecture The interactive real-time interface (IRI) • Design and evaluation results •Conclusion and future work 9/27/2012 3
Background • Electrocardiography is a well known method for heart diagnosis – Used as one of major diagnosis for conventional health monitoring • Electrocardiography main processing challenges arise from : – High computational demand for processing huge amount of data under: • Strict time constraints • Relatively high sampling frequency • Life critical conditions 9/27/2012 4
Background • Most ECG systems use Pan-Tompkins approach based on QRS complex – Usage of R-peak as a reference point – Accurate detection of R-peak is a must • R -peak detection might be inaccurate • Traditional techniques may fail in detecting serious heart problems 9/27/2012 5
Background Heart period detection Heart period R R R mV s 9/27/2012 6
Background Problems of period detection Faulty analysis R- R’ Interval False Interval R-T Interval R R’ False Interval R T mV True Interval True Interval t 9/27/2012 7
Background • The monitoring part is crucial for the real time diagnosis that characterized the ECG signals • Existing Methods : manually-intensive work flow for data acquisition, formatting, and visualization most often relying on multiple serial processes and several software packages 9/27/2012 8
Background :contributions • We propose an efficient ECG heart period detection algorithm, Period Peak Detection (PPD) • A hardware implementation (MPSoC) to deal with the high requirement of biomedical data : real time analysis, high accuracy, portability • An interactive real time interface for the monitoring of the ECG signal targeted for elderly people 9/27/2012 9
Contents • Background • Contributions – Period-Peak Detection (PPD) Algorithm – System architecture System architecture The interactive real-time interface (IRI) •Design and evaluation results •Conclusion and future work 9/27/2012 10
Contents • Background • Contributions – Period-Peak Detection (PPD) Algorithm – System architecture System architecture The interactive real-time interface (IRI) •Design and evaluation results •Conclusion and future work 9/27/2012 11
PPD Algorithm • Based on autocorrelation function (ACF). Peak detection Reading data Derivation Calculate threshold Autocorrelation Find peaks Find interval Store results Period detection 9/27/2012 12
PPD Algorithm- Derivation • Emphasis of the signal peaks • Implementation with simple operations (-) [ 1 ] [ ] y y n y n ( ) t ( 1 ) t n n [ 1 ] [ ] y n y n y [ n ] : current sampling data (filtered orignal ECG signal) , : current ti me (step) t n 9/27/2012 13
PPD Algorithm- Derivation Derivative of the ECG signal Time(sec) Signal peaks P, Q, R, S, T, and U Derivative amplifying R peaks 9/27/2012 14
PPD Algorithm- Autocorrelation • Measures the degree of association between values in a series separated by some lags • Periodicity analysis of signals N [ ] [ ] [ ] R L y n y n L y 0 n R y : the autocorrel ation function y [ n ] : Filtered ECG signals : the number of times needed for the calculatio ns N t o get the period L : the number of lags of the autocorrel ation 15 9/27/2012
PPD Algorithm- Autocorrelation Derivative of the ECG signal The ACF on the derivative AC of the derivative characterized by significant periodic peaks having the Derivative amplifying R peaks same value as the period of the ECG signal 9/27/2012 16
Period detection- Find interval Find maximum value Sort base points Reduce negative value Peak detection from ACF result Calculate interval Renew next start index Find base points 9/27/2012 17
Period detection- Find interval used to determine a threshold Find maximum value Sort base points Reduce negative value Peak detection from ACF result Calculate interval Renew next start index Find base points 9/27/2012 18
Period detection - Find maximum value 9/27/2012 19
Contents • Background • Contributions – Period-Peak Detection (PPD) Algorithm – System architecture System architecture The interactive real-time interface (IRI) •Design and evaluation results •Conclusion and future work 9/27/2012 20
System architecture ADC 1 FIR 1 SDRAM MPSoC ECG Analysis ADC 12 FIR 12 Signal reading Filtering Analysis Server side Client side System architecture IRI 9/27/2012 21
System architecture ADC 1 FIR 1 SDRAM MPSoC ECG Analysis FIR 12 ADC 12 Signal reading Filtering Analysis 9/27/2012 22
Hardware prototyping 9/27/2012 23
Contents • Background • Contributions – Period-Peak Detection (PPD) Algorithm – System architecture System architecture The interactive real-time interface (IRI) •Design and evaluation results •Conclusion and future work 9/27/2012 24
System architecture ADC 1 FIR 1 SDRAM MPSoC ECG Analysis ADC 12 FIR 12 Signal reading Filtering Analysis Server side Client side Hardware prototyping IRI 9/27/2012 25
interactive real-time interface (IRI) Server side Client side 9/27/2012 26
The interactive real-time interface (IRI) 9/27/2012 27
Contents • Background • Contributions – Period-Peak Detection (PPD) Algorithm – System architecture – The interactive real-time interface (IRI) •Design and evaluation results •Conclusion and future work 9/27/2012 28
Hardware Prototyping Results 9/27/2012 29
Performance evaluation 9/27/2012 30
Performance evaluation 9/27/2012 31
Performance evaluation 9/27/2012 32
Contents • Background • Contributions – Period-Peak Detection (PPD) Algorithm – System architecture – The interactive real-time interface (IRI) •Design and evaluation results • Conclusion and future work 9/27/2012 33
Conclusion • Period -Peak Detection (PPD) Algorithm • Scalable Multiprocessor Implementation for ECG Analysis in Multi-lead Records •Sufficient processing speed & 69 % accuracy • Interactive real-time interface (IRI) 9/27/2012 34
Calculation process of ACF • Assumption - The ECG signals are 9 samples. Time t 0 1 2 3 4 5 6 7 8 Signal [ n ] y 0 1 2 0 1 2 0 1 2 8 [ ] [ ] [ ] R L y n y n L y 0 n When , y is L 0 0 n 9/27/2012 35
Calculation process of ACF • Results [ n ] y 0 1 2 0 1 2 0 1 2 L 0 1 2 3 4 5 6 7 8 15 6 4 10 4 2 5 2 0 R y R y L 9/27/2012 36
Calculation process of ACF • Results [ n ] y 0 1 2 0 1 2 0 1 2 L 0 1 2 3 4 5 6 7 8 15 6 4 10 4 2 5 2 0 R y Period Period R y L 9/27/2012 37
Calculation process of ACF • Results [ n ] y 0 1 2 0 1 2 0 1 2 L 0 1 2 3 4 5 6 7 8 15 6 4 10 4 2 5 2 0 R y Period Period R y Every 3 samples are periodic L 38 9/27/2012
Calculation process of ACF (2/6) 8 8 [ 0 ] [ ] [ 0 ] [ ] [ ] R y n y n y n y n y 0 0 n n Signal [ n ] y 0 1 2 0 1 2 0 1 2 × × × × × × × × × Signal [ n ] y 0 1 2 0 1 2 0 1 2 Calculation [ 0 ] 15 R y 9/27/2012 39
Calculation process of ACF (3/6) 8 [ 1 ] [ ] [ 1 ] R y n y n y 0 n [ n ] y 0 1 2 0 1 2 0 1 2 × × × × × × × × [ n ] y 0 1 2 0 1 2 0 1 2 Zero Calculation No calculation (n-L < 0) (n > 8) [ 1 ] 6 R y 9/27/2012 40
Calculation process of ACF (4/6) 8 [ 2 ] [ ] [ 2 ] R y n y n y 0 n [ n ] y 0 1 2 0 1 2 0 1 2 × × × × × × × [ n ] y 0 1 2 0 1 2 0 1 2 Zero Calculation No calculation (n-L < 0) (n > 8) [ 2 ] 4 R y 9/27/2012 41
Calculation process of ACF (5/6) 8 [ 3 ] [ ] [ 3 ] R y n y n y 0 n [ n ] y 0 1 2 0 1 2 0 1 2 × × × × × × [ n ] y 0 1 2 0 1 2 0 1 2 Zero Calculation No calculation (n-L < 0) (n > 8) [ 3 ] 10 R y 9/27/2012 42
Recommend
More recommend
