Gears are critical gearbox components known for their high manufacturing and maintenance costs and operation under harsh conditions. Due to the significant role of gearboxes in mechanical systems, early fault detection is crucial to prevent costly downtime and ensure operational efficiency. This research uses a Convolutional Neural Network (CNN) combined with an Autoencoder (AE) for automatic gear fault classification, aiming to develop an advanced diagnostic tool for identifying gear defects through vibration signal analysis. The data was sourced from publicly available datasets and processed using signal processing techniques in both the time and frequency domains. Key features were extracted using the Wavelet Packet Transform and the CNN-AE model training methods. The experimental results indicate that the proposed model accurately classifies gear faults, surpassing traditional diagnostic methods. This study demonstrates the potential of integrating advanced neural network architectures with signal processing techniques to improve the reliability and accuracy of machine condition monitoring systems.