Creating a test database of images of users' lips in various states to increase the accuracy of the system

Content:

1. Search and analysis of the optimal color space for the construction of eye-catching objects on a given class of images
2. Determination of the dominant signs of classification and the development of a mathematical model of facial expressions "
3. Synthesis of optimal facial recognition algorithm
4. Implementation and testing of facial recognition algorithm
5. Creating a test database of images of users' lips in various states to increase the accuracy of the system
6. Search for the best open source audio speech recognition system
7. Search for the optimal audio system of speech recognition with closed source code, but having open API, for the possibility of integration
8. Experiment for integrating video extensions into audio speech recognition system with test report

Goals:

Increase the accuracy of finding the lips area of ​​the developed system for users with different types of face, skin color, features of the lips, environmental characteristics, lighting and so on.
')

Tasks:

Take photographs of variations in users' lips in various positions to increase the accuracy of the system manually. Implement a system database library for various types of lips in motion based on a free database management system (DBMS).

Introduction

In previous reports, I have analyzed various kinds of color spaces that are optimal for selecting objects in a given class of images. Various algorithms of recognition of a human face and its characteristics were investigated. Also considered are various kinds of mathematical models that can be used to process visual data in real time. On the basis of the conducted research, its own lip-finding algorithm was implemented. However, the accuracy of the active contour model of the object of interest in the presented solution does not always correspond to reality. Since we are faced with the task of implementing a custom product, in order to increase the accuracy of the modular visual solution in this scientific study, we are faced with the task of assembling a library of data on various types of users' lips.

1. Photographing lip variations with different types of faces.

Since, for various reasons, the presented system did not always succeed in correctly localizing the lip area on its own, in order to make the technology more efficient it was decided to investigate different types of users' lips manually in order to enter the initial data into the system database at the next stage.

For the implementation of this testing task, the lip movement of 18 users was analyzed. The source data was photographed manually. Figure 1 shows the faces of users with the location of the lip area during speech.

2. Creating a database of lips system based on MySQL.

MySQL is a free database management system (DBMS). MySQL is owned by Oracle Corporation, which received it together with the absorbed Sun Microsystems, which develops and supports the application [1]. Distributed under the GNU General Public License or under its own commercial license. In addition, developers create functionality for licensed users; it is precisely because of this order that the replication mechanism appeared in almost the earliest versions.

Replication is a mechanism for synchronizing the contents of multiple copies of an object. This process refers to copying data from one source to many others and vice versa, where:
• master — the master server whose data is to be duplicated;
• replica - fixed server that stores a copy of the main data [2].

MySQL is a solution for small and medium applications. Included with the servers WAMP, AppServ, LAMP and portable assemblies of servers Denver, XAMPP. Typically, MySQL is used as a server that is accessed by local or remote clients, but the distribution package includes an internal server library that allows you to include MySQL into standalone programs [3]. It is these libraries that will be used when creating a database on the lips of users.

The flexibility of MySQL DBMS is supported by the support of a large number of table types: users can select both MyISAM type tables that support full-text search and InnoDB tables that support transactions at the level of individual records. Moreover, MySQL DBMS comes with a special type of table EXAMPLE, demonstrating the principles of creating new types of tables. Due to the open architecture and GPL-licensing, new types of tables are constantly appearing in MySQL.

Of course, it makes no sense to put in the database all the information that we have. For example, you do not need to store pictures there, even though MySQL allows it. Placing binary images of graphic files into the database, we will only slow down our server. It is much easier to read a file with a picture from a disk and, from the point of view of consumed resources, it is more economical than connecting from a script to SQL, make a request, get an image, process it and, issuing the necessary headers, show the server visitor. In the second case, the operation of issuing a picture will require several times more CPU, memory and disk resources. It is also worth remembering that there are document caching mechanisms that allow the user to save on traffic, and with dynamic content generation, we actually deprive our visitors of this convenient feature [4].

Instead of images, it is better to store information in MySQL, on the basis of which you can generate links to static images in documents dynamically created by scripts. When implementing our data library, we used indicators derived from changes in the corners of the lips. This point is presented in Figure 3.

The data of statistical changes in the position of users' lips were then collected as a separate database based on the MySQL server DBMS.

Thus, to improve the accuracy of the lip contour finding system, 18 different users were photographed manually in the lip movement database. Photographs of the face and position of the lips of the users were initially analyzed separately. Then, based on the source data, a ready-made library based on MySQL DBMS began to be assembled.

Despite the fact that MySQL DBMS allows you to save photos of users' lips in the system, nevertheless, we decided to abandon this possibility, since this concept complicates the work of the system, since we have to deal with a very large library for processing the visual signal. Therefore, the basis of the work of our system was taken fixing the position of the contours of the lips of users in space. On the basis of this concept, the ability to automatically maintain the static position of users' lips was generated in order to further refine and improve the system.

In the future, further improvement of the database of the system being developed is expected due to the automatic fixation of the lips of users with the most diverse face type, skin color and environmental conditions, as well as the possibility of adjusting the system to the individual characteristics of a particular user.

Bibliography

1) Sample development for MySQL DBMS: macbug.ru/macosxsample/mysql
2) Basics of replication in MySQL: habrahabr.ru/post/56702
3) Plotnikov D.Yu., Malevany EF, Anoprienko A.Ya. Development of a high-loaded web application // Information control systems and computer monitoring (ICS and CM 2012) / Proceedings of the III International Scientific and Technical Conference of Students, Postgraduates and Young Scientists. - Donetsk, DonNTU - 2012, p. 431-435
4) Optimal use of MySQL: www.mysql.ru/docs/optimal.html
To be continued