This study presents a novel technique for translating EEG brain signals into phonemes, which will help people with speech difficulties communicate. The research utilizes a machine learning system that analyzes multi-channel EEG data from 14 electrodes to predict phonemes associated with intended speech. Unlike traditional approaches that require expensive EEG equipment, this project incorporates a budget-friendly simulation framework, combining random signal generation for dynamic authenticity and dataset-driven signal replication for accurate predictions. The methodology involves EEG data preprocessing, feature extraction, and training a fusion model to achieve effective phoneme classification. The findings show considerable accuracy in phoneme prediction, underscoring the potential of EEG-based systems in augmentative and alternative communication (AAC) technologies. Additionally, a simulated hardware prototype and an interactive graphical user interface are created to offer a realistic system demonstration, addressing the limitations of restricted access to EEG hardware. This study fills a critical gap in accessible speech synthesis systems and opens the door to scalable, reasonably priced BCI technology solutions.