Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Improving Pretrained YAMNet for Enhanced Speech Command Detection via Transfer Learning

Sidahmed Lachenani, Hamza Kheddar, Mohamed Ouldzmirli

TL;DR

The paper addresses improving speech command recognition in resource-constrained settings by transferring learning to a pretrained YAMNet model. It processes the Speech Commands subset (32,465 samples across 12 commands) into Mel-spectrogram features and trains a MATLAB-based classifier head, leveraging data augmentation to boost robustness. The method achieves 95.28% accuracy with high precision (95.08%), recall (94.43%), F1 (94.57%), and specificity (99.49%), demonstrating strong performance and generalization. This work highlights the practicality of transfer learning with YAMNet for efficient audio command detection and outlines concrete steps for data conditioning, model configuration, and evaluation, with future directions including deeper architectures and robustness to noise and adversarial threats.

Abstract

This work addresses the need for enhanced accuracy and efficiency in speech command recognition systems, a critical component for improving user interaction in various smart applications. Leveraging the robust pretrained YAMNet model and transfer learning, this study develops a method that significantly improves speech command recognition. We adapt and train a YAMNet deep learning model to effectively detect and interpret speech commands from audio signals. Using the extensively annotated Speech Commands dataset (speech_commands_v0.01), our approach demonstrates the practical application of transfer learning to accurately recognize a predefined set of speech commands. The dataset is meticulously augmented, and features are strategically extracted to boost model performance. As a result, the final model achieved a recognition accuracy of 95.28%, underscoring the impact of advanced machine learning techniques on speech command recognition. This achievement marks substantial progress in audio processing technologies and establishes a new benchmark for future research in the field.

Improving Pretrained YAMNet for Enhanced Speech Command Detection via Transfer Learning

TL;DR

The paper addresses improving speech command recognition in resource-constrained settings by transferring learning to a pretrained YAMNet model. It processes the Speech Commands subset (32,465 samples across 12 commands) into Mel-spectrogram features and trains a MATLAB-based classifier head, leveraging data augmentation to boost robustness. The method achieves 95.28% accuracy with high precision (95.08%), recall (94.43%), F1 (94.57%), and specificity (99.49%), demonstrating strong performance and generalization. This work highlights the practicality of transfer learning with YAMNet for efficient audio command detection and outlines concrete steps for data conditioning, model configuration, and evaluation, with future directions including deeper architectures and robustness to noise and adversarial threats.

Abstract

This work addresses the need for enhanced accuracy and efficiency in speech command recognition systems, a critical component for improving user interaction in various smart applications. Leveraging the robust pretrained YAMNet model and transfer learning, this study develops a method that significantly improves speech command recognition. We adapt and train a YAMNet deep learning model to effectively detect and interpret speech commands from audio signals. Using the extensively annotated Speech Commands dataset (speech_commands_v0.01), our approach demonstrates the practical application of transfer learning to accurately recognize a predefined set of speech commands. The dataset is meticulously augmented, and features are strategically extracted to boost model performance. As a result, the final model achieved a recognition accuracy of 95.28%, underscoring the impact of advanced machine learning techniques on speech command recognition. This achievement marks substantial progress in audio processing technologies and establishes a new benchmark for future research in the field.
Paper Structure (15 sections, 6 equations, 4 figures, 2 tables, 1 algorithm)

This paper contains 15 sections, 6 equations, 4 figures, 2 tables, 1 algorithm.

Table of Contents

  1. Introduction
  2. Related work
  3. Methodology
  4. Dataset Summary
  5. Data Conditioning
  6. Presented Model
  7. Metrics Used
  8. Result
  9. Experimental Setup
  10. Training Configuration
  11. Model Performance Overview
  12. Training Progress Analysis
  13. In-Depth Analysis of Key Metrics
  14. Confusion Matrix Analysis
  15. Conclusion and Future Directions

Figures (4)

  • Figure 1: Number of audio samples per class.
  • Figure 2: Model Processing Workflow.
  • Figure 3: YAMNet training and validation performance metrics progress: (a) Accuracy , (b) Loss.
  • Figure 4: Model’s confusion matrix for speech command.