From a long time ago, human emotion identification from the voice signal has been a research issue in applications involving human-machine interfaces. Emotions are crucial to the mental health of humans. It serves as a vehicle for communicating one's viewpoint or mental condition to others. The process of determining the speaker's emotional state from the speech signal is called speech emotion recognition (SER). Any artificial intelligence system with a small computational power can be taught to recognize few universal emotions such as Neutral, Anger, Happiness, Sadness, etc. The chromogram, the Mel scaled spectrogram, mel-frequency spectrum coefficients (MFCC) are the sources of characteristics that we are retrieving in this work. The emotion or mental health in this study is categorised using a deep neural network. Accuracy is boosted by incorporating the Resnet algorithm.