Durian is one of Thailand's most valuable economic fruits, yet ripeness assessment still relies primarily on experienced inspectors who evaluate the fruit by tapping and listening to its acoustic response. This conventional method is subjective and may lead to inconsistent results. This study presents the development of an AI-based acoustic system for non-destructive durian ripeness classification. The system integrates an acoustic acquisition device with an artificial intelligence model that analyzes tapping sounds. Audio signals are converted into Mel spectrograms and processed using a deep learning model to classify durian into three ripeness levels:

• Unripe
• Mid-ripe
• Ripe

The system provides real-time classification results together with prediction confidence scores to support decision-making. The developed prototype demonstrated rapid, accurate, and consistent ripeness classification, reducing reliance on human expertise while improving the standardization of quality assessment. The proposed innovation offers significant potential for practical implementation in durian orchards, packing houses, and export industries by enhancing quality control, reducing classification errors, and supporting smart agriculture and digital transformation in Thailand's durian supply chain.