Biometric Human Identification based on ECG
T. Lugovaya
Department of Applied Mathematics and Computer Science, Electrotechnical University "LETI", Saint-Petersburg, Russian Federation
This material originally appeared in master's thesis "Biometric human identification based on ECG" in 2005. Please cite this publication when referencing this material. |
Abstract
This research investigates the feasibility of using the electrocardiogram (ECG) as a new biometric for human identification. It is well known that the shapes of the ECG waveforms depend on human heart anatomic features and are different for different persons. But it is unclear whether such differences can be used to identify different individuals. This research demonstrates that it is possible to identify a specific person in a predetermined group using a one-lead ECG. A one-lead ECG is a one-dimensional, low-frequency signal that can be recorded from electrodes on the hands. ECG fragments containing QRS complex, P and T waves extracted from the ECG are processed by principal component analysis and classified using linear discriminant analysis. Using this method on a predetermined group of 90 subjects, the experimental results showed that the rate of correct identification was 96%.
1. Introduction
Biometric technologies are among fast-developing fields of information security, gradually entering into all spheres of human activity. Today only three biometric methods have proved their efficiency, namely, identification based on fingerprints, iris or retina, and face. Hand geometry, voice, writing and typing dynamics, etc. are also useful, depending on the purpose and range of application.
This research aims to develop identification system based on ECG (figure 1). ECG is assumed as an almost unique human characteristic because morphology and amplitudes of registered cardiac complexes are governed by multiple individual factors, in particular, by formation and position of the heart, presence and nature of pathologies, etc.
Figure 1. Example of ECG with agreed notations.
2. The ECG-ID Database
Database contains 310 ECG recordings, obtained from 90 persons:
Collected raw data is rather noisy and contain both high and low frequency noise components. Each recording combine both raw and filtered signals:
The ECG-ID Database itself is available here.
Figure 2. Example of ECG-ID Database record.
3. Identification system
The system feasibility was discussed in [1, 2]. This study suggests other data interpretation and classification techniques, with the system tested on a higher level of live input data. The respective findings are compared below (table 2).
The identification system uses a classical scheme including data preprocessing, formation of input data space, transition to reduced feature space, ECG cycles classification and ECG record identification.
On usability grounds, the study uses brief ECG (10-20 sec) records from a single upper extremity lead (lead I) with sample rate 500Hz and digit 12.
3.1. Data preprocessing
Collected raw data is rather noisy and contain distortions of various origins, both high and low frequency noise components are in presence (figure 3).
A. ECG with power-line noise |
B. ECG with high-frequency noise |
C. ECG with both power-line and high-frequency noise |
D. ECG with isoline drift |
Frequency-selective signal filtering was implemented using a set of adaptive bandstop filter and low-pas filter (figure 4).
Isoline drift correction was implemented using multilevel one-dimensional wavelet analysis. Original signal was decomposed at level 9 using biorthogonal wavelet. Signal reconstructed using final approximation coefficients is assumed to be drifting isoline, which is subtracted from the original signal (figure 5). This method shows good results in both cases of clear and rather noisy ECG signals.
3.2. Initial feature space
The study focuses on ininitial feature space formation. Obviously information on cardiac performance is basically held in the pulse cycle fragment containing the QRS complex and P- and T waves (referred here as the PQRST-fragment). Therefore the stage begins with extraction of a set of R-peak synchronized PQRST-fragments (figure 6). The PQRST fragments length is invariably 0.5 sec or 250 counts.
PQRST-fragment samples are used as informative features. Therefore extracted PQRST-fragments are then processed to enhance their similarity as follows:
1. | Correcting PQRST-fragment mutual "vertical" shift due to eventual residual isoline drift: |
2. | Culling eventual "atypical" PQRST-fragments due to gestures, deep breathing or certain pathologies: |
3. | Correcting PQRST-fragments depending on heart rate using Bazett's formula: |
|
Thus in the initial feature space (dimension N=250) ECG appears as a set of PQRST-fragments with each seen as a separate pattern at subsequent system stages, to be interpreted and classified independently.
3.3. Feature space reduction
The feature space is reduced using Principal Components Analysis (PCA) so that space dimension can be reduced to 30 (with the Kaiser criterion) or even to 10 (with the scree test).
Or, alternately, the space is reduced with wavelet transform (WT) providing the same space reduction but with slightly poorer final identification.
3.4. Classification and Identification
The resulting PQRST-fragment patterns are then classified in reduced feature space using linear discriminant analysis.
At the final stage, the ECG record identification is based on PQRST-fragments classification results.
4. Results
Experimental studies involve 90 human. ECG records were made in the sitting position, heart rate, physical and emotional state were not limited. Collected data set contains 320 records, of which 200 records were assigned to the training set and 120 records to test set. Differentiation between trainig and test sets aimed to provide for maximum performance complexity, i.e. maximum difference between records in different sets both in monitoring time and human physical state.
Averaged results of series of experiments on PQRST-fragments classification and ECG record identification with different feature space reduction methods are tabulated in Table 1. ECG identification leveling is thus 96%.
Reduction technique | Number of features | PQRST-fragments classification, % | ECG identification, % | |
Training set | Test set | Test set | ||
PCA | 10 | 99 | 85 | 89 |
WT | 9 | 98 | 79 | 82 |
PCA | 30 | 99 | 91 | 96 |
WT | 34 | 99 | 88 | 91 |
Additionally, results from [1, 2] and this research are compared in Table 2.
Research | Class count | Identification results, % |
[1] | 20 | 98 |
[2] | 9 | 95 |
this | 90 | 96 |
As a result of this research a recognition system was developed, it solved the problem of biometric human identification based on ECG on sufficiently large set of input data. The findings represent primary arguments for ECG usability as a biometric characteristic in various biometric access control problems, opening up brand new perspective for the study of biometric technologies, and extending potentialities of security- and modern amenity systems.
References
[1] L.Biel, O.Pettersson, L.Philipson, P.Wide "ECG Analysis: A New Approach in Human Identification", IEEE Transactions on Instrumentation and Measurement, vol.50, N3, June 2001, pp. 808-812.
[2] W.J. Yi, K.S. Park, D.U. Jeong "Personal Identification From ECG Measured Without Body Surface Electrodes Using Probabilistic Neural Networks".