Function and speed portability of audio fingerprint extraction across computing platforms

Abstract

Audio Fingerprinting (AFP) is a technology, which requests huge computing power for responsiveness, accuracy, and robustness to noise. In this study, we make efforts to improve the computing speed of fingerprint extraction in AFP system by parallelism language OpenCL. Especially, we also explore the function and speed portability across different platform. The experimental results show that the portability of function is trivial by syntax check and utilizes conservative syntax for different compilers, such as Microsoft Visual Studio and Linux GCC. On the other hand, the speed portability requests much works for optimization of different sub modules for the specified platform.