I am highlighting the work done by me in the area of Phonology analysis from childrens’ speech. To start the work main hurdle was availability of the database of processes in children of age from 2.60 years to 6.60 years. I visited many schools and houses to record the database from children. After database collection, I analyzed the nature of mispronunciation in specified age group and categorized them into different types of phonological processes. Each phonological process have specific mispronunciation pattern. Spectral, prosodic, excitation source and signal level properties of the phonological processes are explored to identify the features which help to discriminate the pronunciation errors. With the help of identified reliable features, identification of the mispronounced region and phonological processes has been performed using DTW comparison. I employed speech recognition (Hidden Markov Model) based approach for mispronunciation detection which recognizes the mispronunciation using various Goodness of Pronunciation features (posterior probability, log-likelihood scores, etc.). During the course of implementation, we found an interesting problem of children gender identification. I build a Deep Neural Network model and Random forest ensemble learning model for children gender identification. Emotion recognition using Spectral features using Random Forests. During my five years of PhD, I worked on multidisciplinary research areas such as Speech Processing, Image Processing, Medical imaging. In these fields, I have mostly worked with the various machine learning algorithms such as Random Forests, Support Vector Machines, Artificial Neural Networks, Convolutional Neural Networks, etc.