Achieving communication security, along with high computational efficiency, is one of the challenging issues in the advancement of modern resource constraint wireless networks. Wireless physical layer secure key extraction in conjunction with suitable preprocessing techniques may be a possible way out. Principal component analysis (PCA) is one of the dimensionality reduction techniques employed commonly in various domains for different applications. However, the physical layer secure key extraction employing PCA as dimensionality reduction is untouched so far. This paper presents a comprehensive study on PCA based wireless secret key extraction with real-time experimentation. In this work, we propose to apply PCA as a preprocessing technique to reduce the total number of numerical computations required in the key generation process, by cutting down the dimension of the input data set. We propose to select the extracted principal components to be processed further for key generation, based on their information content and cross-correlation. We analyzed the performance of the proposed in terms of bit disagreement rate, key randomness and pass ratio. The computational complexity of the proposed approach is derived and the effect of dimensionality reduction factor () on the required numerical computations is analyzed. It is found that substantial improvement in bit disagreement performance is achieved along with a significant reduction in the required numerical computations. Remarkably, these outcomes are achieved by slightly modifying one of the blocks of the traditional key generation system. Furthermore, the practicability of the proposed technique is verified through real-time experimentation in different physical scenarios.