In the fast growing scientific world, the social networks and entertainment have become an essential evil and contain a variety of face images. The face images have large visual variations like variety of pose, lighting, and variation in expression, aging, alignment and occlusion. All these factors act in performance degradation for face recognition and analysis. The face analysis is a sequence of face detection, face alignment and face recognition. The face alignment problem is scarcely studied in the past. Convolutional neural networks (CNNs) have been used for the facial keypoint detection. Alignment variation is a challenging task in face recognition. Therefore, in the present work, a novel feature extractor method is developed by the combination of Hough transform and CNN defining face features matrix dimensions. In our work, we have used following parameters such as left eye (LE), right eye (RE), nose tip (N), mouth (M). The technique is useful with accuracy of detection and alignment ranging from 90% to 100% and improved precision.