One of the significant stages in a gas identification system is dimensionality reduction to speed up the processing part. This is even more important when the system is implemented on a hardware platform where the resources are limited. This paper presents the design and the implementation of the learning and testing phases of principal component analysis (PCA) that can be used in a gas identification system on the heterogeneous Zynq platform. All steps of PCA starting from the mean computation to the projection of data onto the new space, passing by the normalization process, covariance matrix and the eigenvectors computation are developed in C and synthesized using the new Xilinx VIVADO high level synthesis (HLS). The computation of the eigenvectors was based on the iterative Jacobi method. The designed hardware for computing the learning part of PCA on the Zynq system on chip showed that it can be faster than its 64-bit Intel i7-3770 processor counterpart with a speed up of 1.41. Optimization techniques using HLS directives were also utilised in the hardware implementation of the testing part of the PCA to speed up the design and reduce its latency.
|Title of host publication||2013 IEEE 20TH INTERNATIONAL CONFERENCE ON ELECTRONICS, CIRCUITS, AND SYSTEMS (ICECS)|
|Publication status||Published - 2013|
|Name||IEEE International Conference on Electronics, Circuits and Systems|