Emotune: Audio Emotion Analyzer

• Trained a 1D‑CNN on the RAVDESS dataset (2,440 audio clips across 8 emotions), extracting 40 MFCC features per sample • Achieved 94.8% accuracy and 0.95 macro F1‑score on the test set after 50 epochs, reducing preprocessing time by 20% through optimized pipelines • Implemented dropout (0.2) and max-pooling layers to prevent overfitting, boosting minority-class F1 by 12% • Visualized performance with confusion matrix and classification report, ensuring > 90% recall across all emotion classes