Trigger word detection model

After obtaining my Deep Learning certification from DeepLearning.AI, I put my new skills into practice by embarking on an ambitious project. Thus, I developed a trigger word detection system, named โ€œ๐“๐ซ๐ข๐ ๐ ๐ž๐ซ ๐–๐จ๐ซ๐ ๐ƒ๐ž๐ญ๐ž๐œ๐ญ๐ข๐จ๐งโ€.

This project aimed to identify, in MP3 audio files, the presence of a specific word pronounced by the speaker. I chose to focus on the word โ€œ๐›๐จ๐ฒโ€.

Thanks to a GRU-based model (Gated Recurrent Unit) that I trained, I achieved a remarkable accuracy of 95.50%.

I invite you to explore the GitHub repository: GitHub - Mouaadag/trigger-words-detection: Build a model with tensorflow to trigger a word on a mp3 files. , where I detail the steps of data preparation, feature extraction, and the construction and training of the model.

I would like to express my gratitude to the DeepLearning.AI team for the exceptional quality of their training, and a special thanks to Andrew Ng for his courses and his clear and inspiring way of conveying Deep Learning concepts.

7 Likes

Cool project! And wow, 95% accuracy! What are your plans for future projects? Greetings, Julius from Germany. (A beginner in AI, about to finish Deep Learning course 1)