Acquisition of heartbeat signal from facial video and its application in biometric authentication

Start Date: 01/02/2015

End Date: 28/02/2015

Applicant: Mr Mohammad Ahsanul Haque

1. Purpose of the STSM

Title: Acquisition of heartbeat signal from facial video and its application in biometric authentication

Summary: Now-a-days the utilization of heartbeat signal as a biometric trait is an emerging area of study. Present methods use Electrocardiogram (ECG) to obtain heartbeat signal. Use of an ECG requires the user to wear electrodes or chest straps, which are obtrusive, and fail in covert and unintended identification scenarios and where users’ skin is sensitive to the sensors. On the other hand, recent advances in Heartbeat Rate (HR) measurements from facial video provide an insight of the relationship between HR, and subtle facial skin color change. Facial video based HR measurement methods work by extracting trajectories of facial skin color change and then employing signal processing method on the trajectories to measure HR. Thus, the purpose of the STSM included the works regarding to the following research themes:

  • Defining the relationship between ECG signal, and facial skin color change.
  • Extracting heartbeat signal from facial video.
  • Extracting relevant and effective features of the heartbeat signal in order to use these for biometric authentication.

 

2. Description of the work carried out during the STSM

The intended works have been carried out during the STSM with the collaboration with another PhD student under the supervision of Prof. Abdenour Hadid in the Center for Machine Vision Research (CMV), University of Oulu, Finland. The following works have been carried out:

  • The characteristics of ECG signal and facial skin color change have been studied from the literature in order to understand and confirm their correlation.
  • A system, including face detection, Region of Interest (ROI) detection and facial skin color change tracking, was developed to extract the heartbeat signal from facial video.
  • A system was developed by using signal processing methods such as filtering, denoising and Radon transformation to extract features from the heartbeat signal.
  • Face recognition by the extracted features by using a Support Vector Machine
  • Employing the above mentioned system on a publicly available video database for face recognition.

 

3. Description of the main results obtained

We employed our system on a large publicly available video database, called MOBIO. This has around 40 subjects with more than 10000 videos.

The experimental results shows that the heartbeat signal from facial video can discriminate between different person’s face with an accuracy range of 5-10% on different subset (evaluation and training) of MOBIO database.

 

4. Future collaboration with the host institution (if applicable)

During the STSM another investigation were made along with the investigation of the heartbeat signal from facial video by collaboration with another PhD student at CMV. We implemented the Local Binary Pattern (LBP) and its variant LBP-TOP based face recognition system and employed it on the same database of the aforementioned experiment.

Our plan is to fuse the LBP and heartbeat from facial video and examine the results of face recognition in MOBIO database. We have agreed upon the fusion level, and it is ‘score level fusion’.

 

Comments are closed.