Human Pose Recognition (HPR), the work of estimating the spatial configuration of human's body through a video has shown notable recognition in deep learning and artificial intelligence communities. HPR is a demanding technology that is used across numerous applications to estimate human pose actions in real-time.HPR is the process of identifying the activities being performed by human through analysing video data traces of their movements. Our proposed methodology uses real world dataset to draw precise labels with the probabilities for corresponding human poses. The primary aim of our proposed methodology is to label activities by analysing video data. This is achieved initially by taking real-time video input from the user using OpenCV and performs pose recognition. In our proposed methodology, we used six different categories such as sitting, standing, walking, pain, falling, sleeping using a Mediapipe 32-point landmark key-point detection model and hybrid CNN-LSTM model. The CNN-LSTM enhances pose recognition by effectively capturing both spatial and temporal features. However, probabilities of poses such as sitting, standing, walking, pain, falling, sleeping are analysed and achieved overall 96.5% accuracy.