Human Speech Processing is one of the most important areas in Digital Signal Processing and is used in Human Computer Interface, Telecommunications, Assistive Technology, Audio Production, Security, etc. to perform many operations. Human Speech Emotions Recognition (HSER) is a technique that can be utilized to detect emotions reflected in various speech patterns. The task is similar to text sentiment analysis, differing only in the data modalities. From the Machine learning perspective, it can be treated as a classification problem where input speech patterns must be grouped into some already defined emotions. Multiple factors are to be considered while working with speech data, such as speaker, language, region, and emotions, to name a few, that make the problem of HSER complex. The complexity increases multifold when the input speech patterns are ambiguous, making it challenging to apply HSER in real-world. The current paper proposes a novel Convolutional Neural Network (CNN) based HSER framework to detect emotions in human speech effectively. It is evaluated on the popular RAVDESS dataset. The results received are further compared with the MLP classifier on accuracy.