Actually, the design of the portable facial recognition system be separated into five major categories which include:

1. Detection of Face

2. Face Preprocessing

3. Training

4. Recognition of Face

5. Loading and Saving the Data

Following is the block diagram which has been developed for the code:

Block Diagram

In accordance with the above figure, it can be observed that the green block is illustrating the mode of detection. Generally, this code part is running without any activation or initialization from the user. It means that this process is operating by default and doesn’t need to be activated by the user. First of all, a new look and image of the face and eyes are acquired by it. And with the detection of eyes and face, preprocessing is carried out on the region of face which includes a face. In this manner, the image is then utilized in the next block. Moving on, the next block is triggered which is purple. It doesn’t operate by default and it is triggered normally when the button of Add Person is clicked by the user. This process will work to add faces which are preprocessed for the corresponding person. And once it is indicated by the user that they have completed the process of collecting faces, the code will continue to the yellow block which is the Training Mode(Meenakshi, 2013).

In this manner, the code will not break and will initiate the yellow block. In this block, the algorithm model of Fisher faces is trained through the use of preprocessed faces. In addition to these faces, associated labels are also used. And once training has is completed, the code transitions automatically to the red block which is Recognition Mode. It can be said that in the mode of recognition, the program will check if the faces which are captured seem to match the trained faces. Upon matching, personalized music will be played by the system. If the person is not recognized then the code will simply go back to the acquisition of new image. In addition to it, there will be no special action which will be taken.

Some of the side tasks include the capability of saving faces so that it is not necessary for users to train new data sets each and every time. This seems to be related directly with the loading functionality of faces being the function of loading the faces which have been saved previously. All the faces can also be deleted by users if they need to start the process of training again. Considering the fact that side tasks are based on mouse click, they don’t follow the same processes which are required in main functions. In the figure below, the setup of system can be observed. In the later sections of this paper, more specific detail will be provided(Introna & Nissenbaum).

A = Light fixture, G = Keyboard, B = Webcam, F = Mouse, E = Speaker, D = Raspberry Pi, and C = Display.

 Hardware

In the system of portable facial recognition, the computer uses is Raspberry Pi. When it comes to its core, it has a quad-core of 900MHz and it has a VideoCore IV which possesses a 250 MHz GPU. Additionally, it also includes a RAM of 1GB and the reason why Raspberry Pi is utilized is that it is quite affordable and also includes sufficient processing power for algorithms operating in the facial recognition. Other than these specifications, HD Webcam C270 of Logitech was utilized as the device to capture images since it was compatible with UVC. It means that has the capability of streaming video with an interface of USB.

The remaining components seem to have flexibility such as light fixture, speaker, keyboard, and mouse. The keyboard and mouse are utilized for having an interface with the Linux Operating System of the Raspberry Pi for clicking, and for clicking on buttons and typing commands for executing different operations. The music file is operated by the speaker. Lastly, the light fixture is utilized for controlling the lighting in such a way that there is consistent and strong light on the face as the test is performed.

Software Used of Portable Low-Cost Facial Recognition

In order to write the source code, the language of C++ was utilized as it is quite a fast computer language in comparison with MATLAB. The library of OpenCV was also utilized for the functions of computer vision and the library of SDL or Simple DirectMedia Layer was utilized for playing the sound. Both of the libraries can be used on different platforms which means that they are cross-platforms and are compatible with both Windows and Linus Operating Systems (Kaur, Vashisht, & Neeru, 2010).

Detection Mode of Portable Low-Cost Facial Recognition

In the mode of detection, the program attempts to determine or identify if a face exists within the frame captured by the system or not. For eyes and frontal face, there are several training models available in the website of OpenCV. In addition to it, LBP or Local Binary Patterns classified is utilized in this system for identifying the face because it is normally faster than the classifier or Haar. This rapid rate makes it better for different real-time operations and processes.

In comparison with Haar, LBP is less precise but then, we will need the detection and identification of both eyes within the region of face for added accuracy and reliability. With the detection of face, the program will be looking for eyes within the region of face to be utilized in the section of processing. It is important for both of the eyes to be detected for preprocessing to be precise. There is a significant likelihood that issues might occur if the detector of eye is utilized over the region of face. If issues occur, the solution to the misidentification is to specify the top right and left face regions for finding the right and left eye. Generally, these regions are based on different geometric restrictions in which most of the human eyes will be located with respect to the facial structure. And reduction in the region of eyes serves to decrease the processing time while increasing the precision of eye detection.

References of Design of Portable Low-Cost Facial Recognition

Bailey, J. (2018). Data protection management in UK library and information services. iConference 2018 Proceedings.

Floridi, L., & Taddeo, M. (2016). What is data ethics?

Introna, L., & Nissenbaum, H. (n.d.). Facial recognition technology a survey of policy and implementation issues. 2010.

Kaur, M., Vashisht, R., & Neeru, N. (2010). Recognition of facial expressions with principal component analysis and singular value decomposition. International Journal of Computer Applications, 9(12), 36-40.

Meenakshi, M. (2013). Real-Time Facial Recognition System—Design, Implementation and Validation. Journal of Signal Processing Theory and Applications, 1, 1-18.