Incorporating Noise Robustness in Speech Command Recognition by Noise Augmentation of Training Data

Area of Science:

• Computer Science
• Artificial Intelligence
• Signal Processing

Background:

• Speech recognition technology has advanced significantly due to new devices, machine learning, and large speech corpora.
• Speech command systems are crucial for robotics, IoT, and human-computer interfaces, but noise significantly degrades their performance.
• Existing research has not adequately addressed the challenge of real-time noise in speech command datasets.

Purpose of the Study:

• To analyze current trends in speech recognition.
• To evaluate existing machine learning and deep learning techniques on speech command datasets.
• To propose and validate a novel technique for enhancing noise robustness in speech recognition systems.

Main Methods:

• Thorough analysis of the latest speech recognition trends.
• Evaluation of speech command datasets using various machine learning and deep learning models.
• Development and implementation of a novel noise augmentation technique for training data.

Main Results:

• The proposed noise augmentation technique significantly improves speech recognition performance.
• The technique demonstrates superior results on both clean and noisy datasets compared to state-of-the-art methods.
• A new benchmark for noise-robust speech recognition has been established.

Conclusions:

• Noise augmentation in training data is an effective strategy for improving speech recognition robustness.
• The developed technique offers a significant advancement in handling real-world noisy conditions for speech command systems.
• This research paves the way for more reliable and efficient speech-enabled applications in diverse environments.