In an increasingly aged society, demand for hearing aids is growing due to age-related hearing loss. However, adoption remains low, especially among older adults with mild hearing loss, who often avoid hearing aids due to cost, discomfort, and limited improvement in speech intelligibility. To address this, we propose a system that enhances intelligibility by emphasizing consonants based on their phoneme categories, such as fricatives, plosives, and nasals. This study focuses on phoneme category classification as a core component. We implement and evaluate classifiers using convolutional neural networks (CNN) and long short-term memory with fully convolutional networks (LSTM-FCN). Training data were drawn from two large-scale Japanese speech corpora, and multiple segmentation strategies were tested. Results show that CNN outperforms LSTM-FCN in classification accuracy, and the approach remains robust even with ambient noise. This paper outlines the system, classification framework, and experimental results, demonstrating the feasibility of phoneme-category-based enhancement and its potential for real-time, personalized auditory support.