Selection of appropriate data storage for wavelet transform coefficients

Abstract

The study of transform coefficients brought a question of a useful tool for signal evaluation and comparison. The application should use a database storage system for fast inserting and selecting data and should be also capable of saving large amount of real numbers. This paper deals with performance tests of selected database management systems: relational MySQL, MS SQL and object-relational PostgreSQL. The tests were focused on different storage engines and used data types such as a numerical array, text and decimal. Besides the relational database models, a multidimensional approach was projected and analyzed. The results showed that the most effective solution is saving the transform coefficients into separated 16 atributes using the MySQL storage engine called MyISAM. The multidimensional model appeared to be less suitable for this kind of application. © 2013 Springer Science+Business Media.