Elsevier

Pattern Recognition

Volume 41, Issue 3, March 2008, Pages 906-919
Pattern Recognition

Heart sound as a biometric

https://doi.org/10.1016/j.patcog.2007.07.018Get rights and content

Abstract

In this paper, we propose a novel biometric method based on heart sound signals. The biometric system comprises an electronic stethoscope, a computer equipped with a sound card and the software application. Our approach consists of a robust feature extraction scheme which is based on cepstral analysis with a specified configuration, combined with Gaussian mixture modeling. Experiments have been conducted to determine the relationship between various parameters in our proposed scheme. It has been demonstrated that heart sounds should be processed within segments of 0.5 s and using the full resolution in frequency domain. Also, higher order cepstral coefficients that carry information on the excitation proved to be useful. A preliminary test of 128 heart sounds from 128 participants was collected to evaluate the uniqueness of the heart sounds. The HTK toolkit produces a 99% recognition rate with only one mismatch. Next, a more comprehensive test consisting almost 1000 heart sounds collected from 10 individuals over a period of 2 months yields a promising matching accuracy of 96% using the proposed feature and classification algorithm. A real-time heart sound authentication system is then built and can be used in two modes: to identify a particular individual or to verify an individual's claimed identity.

Introduction

In recent years, the use of a reliable authentication and identification system to identify legitimate user is becoming increasingly important in commercial application, personnel security, military, finance, airport, hospital, digital right management systems and many other important areas [1]. In fact, performance-based biometric systems whereby a person is automatically recognized by him performing a pre-defined task using his own biometrics, are preferred over knowledge-based (e.g., password) or possession-based (e.g., key) access control methods. As a result, conventional biometrics systems like fingerprint, iris, face and voice that provide recognition based on an individual behavioral and/or physiological characteristics are becoming more popular [1], [2], [3], [4]. However, a common weakness of these system is their vulnerability to the possibility to falsify these features [2], [3], [5], [10], [11].

For increased reliability and added security, one approach is the use of multimodal biometric systems which uses multiple biometric modalities (such as face and iris of a person or multiple fingers of a person) [2]. New biometrics such as hand vascular pattern, vein, gait, human tissue, knuckle, ear canal and even evoked brain signals have also been proposed [6], [7]. Current works on these areas are mainly focused on increasing the reliability and convenience of data capturing, as well as improving the system accuracy and robustness.

Beside these, the study of using the electrocardiogram (ECG) [8], [9] as a biometric has also been carried out, which yields a relative high result for human identification tasks [8], [9]. In Ref. [8], ECG measurements are collected from 20 male and female adults over a six week period. The training and testing set consist of 85 and 50 measurement sets, respectively, and produce a 98% recognition rate. Israel et al. [9] further investigated the effect of the state of anxiety of an individual on its ECG features through a series of high and low stress tasks. Test results show that the features extracted from the ECG signal are unique to an individual and invariant to the individual state of anxiety. Israel et al. [9] also found that the identification performance is independent of the electrode placements. However, we note that ECG for identification is generally cumbersome due to the many (at least three) electrodes required [8].

In this work, we investigate the possibility of using human heart sounds—an acoustic signal—as a reliable biometric for human identification based on the following requirements mentioned in Ref. [2]:

  • (1)

    Universal: Each living human being has a heart that keeps on pumping until his death.

  • (2)

    Quantifiable: Heart sound can be digitally captured and measured using an electronic stethoscope.

  • (3)

    Vulnerability: While there are many biometric systems that are commercially affordable and reliable in the market, most of these biometrics can still be forged by a determined and trained perpetrator as demonstrated in some Hollywood thriller. In other words, these stable biometric which provide static information about the user can be copied and reproduced to fool the biometrics system [2], [3], [5] with the intention to commit fraud. Table 1 [10], [11] summarizes the vulnerability of some commonly used biometric.

     Unlike other biometric technologies that use fingerprints, pictures or static bio-signals to identify a person, heart sound cannot be copied and reproduced easily as it is based on intrinsic human biometric dynamic signals acquired from the body. When compared with voice, we feel that the heart sound exhibits two significant advantages. First, human voice can be easily obtained using a concealed recorder without the person knowing whereas to acquire a person heart sound, the recording device has to be placed on the person chest surface, with his permission, to capture the vibrations from the heart. Second, to reproduce the same heart sound of an individual, an artificial pumping heart with the same anatomy needs to be reconstructed; the physical attribute surrounding the artificial heart must also match the same individual body structure. Hence, heart sound cannot be easily recorded and simulated accurately.

  • (4)

    Acceptability: Privacy issue will affect the extent to which people are willing to accept the use of their heart sound as a biometric identifier in their daily lives [12]. As the auscultation of heart sound can be used as a diagnosis to heart disease, there are fears that these sensitive medical information might be abused unethically by some who may deny benefits to a person determined to be of high risk. To alleviate these fears, legislation might be necessary to ensure that such information remains private and that its misuse will be punished. Alternatively, instead of storing the sensed physical characteristics of the original heart sound waveform, the biometric systems can store a digital representation of the heart signal (in the form of a template). Hence, privacy is safeguarded as the actual physical characteristic cannot be recovered from the digital template. Encryption can be added to ensure that only the designated application can use this template.

  • (5)

    Usability: With the advancement in wearable computing, wireless sensors can eventually be placed on human to capture their heart signal when they are moving or performing other activities. These signals are fed wirelessly to a remote biometric recognition system which will determine the accessibility of the person. As a result, authentication can be done even before the user reaches the gate, resulting in time saving.

     The following three factors will also be validated experimentally in the next section.

  • (6)

    Unique: The physical state of an individual's health, age, size, weight, height, structure of the heart as well as the genetics factors all contribute to an individual's unique heart sound. The heart sounds of two persons having the same type of heart diseases also vary.

  • (7)

    Variability: In a controlled environment, the human heart sound remains sufficiently invariant over a specific period of time.

  • (8)

    Performance: The achievable recognition accuracy and speed of the recognition system will determine it's applicability. A real-time implementation of the proposed biometric system will be discussed in the later section.

Human heart sounds are natural signals, which have been applied in the doctor's auscultation for health monitoring and diagnosis for thousands of years. In the past, study of heart sounds focus mainly on the heart rate variability [13]. However, we conjecture that since the heart sounds also contain information about an individual's physiology, such signals have the potential to provide a unique identity for each person. Like ECG, these signals are difficult to disguise and therefore reduces falsification. Moreover, heart sounds are relatively easy to obtain, by placing a conventional stethoscope on the chest, for example.

In Section 2 of this paper, we will provide the problem formulation. The use of heart sound as a biometric will be evaluated on a number of assessment criteria in Section 3, with various experiments demonstrating the effectiveness and accuracy of the proposed scheme. Next in Section 4, we will introduce the proposed biometric system and provide a detailed description of its various operations. Finally, in Section 5 of this paper, we will conclude our findings.

Section snippets

Problem formulation

In this section, we will briefly outline the mechanisms involved in heart sound production. Next, we will elaborate on the feature extraction and pattern matching scheme used.

Experimental results and discussions

The heart sounds used in our experiments were recorded using a Welch Allyn Meditron electronic stethoscope, as shown in Fig. 7. The electronic stethoscope was placed on the chest of the participant seated in a relaxed position. The heart signal was captured using the CoolEdit software application via the sound card of the computer with a sampling rate of 2 KHz and 16 bits. A Pentium IV 2.4 GHz IntelR personal computer with 512M ram running MicrosoftR WindowsR XP operating systems is used.

First,

Proposed biometric system

In the previous sections, we have verified that the human heart sound can indeed be used as a biometric. Here, we will introduce the proposed biometric system and provide a detailed description of its various operations. In 4.1, the five functional modules that make up the proposed system are described. 4.2 describes how we can possibly implement the various modes of the proposed biometric system. Finally in 4.3, a real-time implementation of the proposed system is realized and detailed.

Conclusions

In this paper, we have investigated the possibility of using heart sound in the human identification task. After a preliminary study on human heart sound signal, we can cautiously comment that the heart sound can be used as a potential new biometrics since it is generally acceptable, and is sufficiently robust to various fraudulent methods and attacks to the system. The heart sound can be used alone or combined with other biometrics like fingerprint, iris, face and/or voice to achieve a more

About the Author — KOKSOON PHUA received his Bachelor and Master degree in Electrical and Electronic Engineering from Nanyang Technological University, Singapore, in 1997 and 1999, respectively. In the year 2003 and 2004, he has received merit recognition for his works in Tan Kah Kee Young Inventors’ Awards. Currently, he is with the Institute for Infocomm Research, A*STAR in Singapore, working on Non-invasive Brain Computer Interface using Electroencephalography (EEG) signal. His research

References (20)

  • S.A. Israel et al.

    ECG to identify individuals

    Pattern Recognition

    (2005)
  • J. Ortega-Garcia et al.

    Authentication gets personal with biometrics

    IEEE Signal Process. Mag.

    (2004)
  • A.K. Jain et al.

    An introduction to biometric recognition

    IEEE Trans. Circuits Syst. Video Technol.

    (2004)
  • L.O. Gorman

    Comparing passwords, tokens, and biometrics for user authentication

    Proc. IEEE

    (2003)
  • J.P. Campbell

    Speaker recognition: a tutorial

    Proc. IEEE

    (1997)
  • M. Faundez-Zanuy

    On the vulnerability of biometric security systems

    IEEE Aerosp. Electron. Syst. Mag.

    (2004)
  • D. Peter

    Biometrics continue to evolve

    Biom. Technol. Today

    (2003)
  • R. Palaniappan et al.

    Individual identification technique using visual evoked potential signals

    IEE Electron. Lett.

    (2002)
  • L. Biel et al.

    ECG analysis: a new approach in human identification

    IEEE Trans. Instrum. Meas.

    (2001)
  • F. Dario

    Biometrics: future abuses

    Comput. Fraud Secur.

    (2003)
There are more references available in the full text version of this article.

Cited by (118)

  • ECGsound for human identification

    2022, Biomedical Signal Processing and Control
    Citation Excerpt :

    Some authors have explored the use of cardiac sounds, specifically phonocardiograms (PCGs), for biometrics purposes in the cybersecurity context [47]. The majority of these proposals extract handcrafted features on a transform domain [48–50]. In this regard, in [51], we can find a comparative study of several cepstral features to characterise PCG records for human recognition.

  • PCG signals for biometric authentication systems: An in-depth review

    2021, Computer Science Review
    Citation Excerpt :

    VQ gives the highest performance and lowest computational time [20,33], and [40]. Some studies used GMM as a classification method [20,24,25,27,35], and [39]. GMM is considered the generalization of the VQ.

  • Acoustic-pressure sensor array system for cardiac-sound acquisition

    2021, Biomedical Signal Processing and Control
View all citing articles on Scopus

About the Author — KOKSOON PHUA received his Bachelor and Master degree in Electrical and Electronic Engineering from Nanyang Technological University, Singapore, in 1997 and 1999, respectively. In the year 2003 and 2004, he has received merit recognition for his works in Tan Kah Kee Young Inventors’ Awards. Currently, he is with the Institute for Infocomm Research, A*STAR in Singapore, working on Non-invasive Brain Computer Interface using Electroencephalography (EEG) signal. His research interests include pattern recognition, machine learning, signal processing, speech enhancement, noise cancellation, and microphone array technique.

About the Author — JIANFENG CHEN received his B.E., M. Eng., and Ph.D. degrees in Electronic Engineering, from Northwestern Polytechnical University, China, in 1993, 1996, and 1999, respectively. In 1999, he joined School of EEE, Nanyang Technological University, as a research fellow. From 2001, he was with Center for Signal Processing under NSTB, Singapore. He is currently a Senior Research Scientist in Institute for Infocomm Research, A*STAR, Singapore. Dr. Chen is experienced in the fields of electrical engineering with strong R&D background in acoustic array processing, speech processing, biomedical signal processing, modern spectral analysis, and related areas. He has more than 10 years teaching experiences in signal processing, circuit and system, and sensor technology. Dr. Chen has published more than 40 papers in the international journals and conferences. He is also the first inventor of three international patents. As the team leader, he received Tan Kar Kee Young Inventor Awards in 2003 and 2004, respectively. His current research interests include acoustic array processing, multimedia signal processing, biomedical signal processing, and wireless sensor network. Dr. Chen is an IEEEE member, Pattern Recognition Society member, and International Communication Society member.

About the Author — TRAN HUY DAT was born in Hanoi, Vietnam, in 1971. He received his Masters degree of Engineering Science in 1995 from the Ukrainian National Technical University. In 2000, he received his Ph.D. degree of Physic-Mathematical Science from the National Academy of Sciences of Ukraine. From 2000 to 2002 he did his postdoc research at the Institute of Hydromechanics, National Academy of Science of Ukraine. From 2002 to 2005 he was a postdoc fellow at the Itakura and Takeda Labs, Nagoya University, Japan. Since 2005 he has been a Senior Research Fellow at the Institute for Infocomm Research, Singapore. His research interest is including acoustic and speech signal processing, neural signal processing and machine learning. He served as a reviewer of several international journals and conferences, including the IEEE Transaction on Audio, Speech and Language Processing and the Neurocomputing.

About the Author — LOUIS SHUE received his Ph.D. from the Australian National University in 1999. After that he joined the Centre for Signal Processing, NTU, where he was a key member for the 2 12-year Multi Algorithm Infra-Red Search and Track (MAIRST) project on image-based tracking of small targets, collaborating with the Defense Science Organization (DSO) of Singapore. After joining the Institute for Infocomm Research, he has led a research group in the field of acoustic signal processing. He has received recognition for Tan Kah Kee Young Inventors’ Awards for “Passive Fetal Heart Monitoring System using Bio-Acoustic Sensor” (Merit Award) in 2004 and for “Accoustic System for Bathroom behaviour understanding of Dementia Patients” (Commendation Award) in 2005. His current research interests are health and telemedicine.

View full text