Samsung electronics co., ltd. (20240355320). SYSTEMS AND METHODS FOR TRAINING ARTIFICIAL NEURAL NETWORKS simplified abstract
Contents
SYSTEMS AND METHODS FOR TRAINING ARTIFICIAL NEURAL NETWORKS
Organization Name
Inventor(s)
Behnam Babagholami Mohamadabadi of Escondido CA (US)
Mostafa El-khamy of San Diego CA (US)
Kee-Bong Song of San Diego CA (US)
SYSTEMS AND METHODS FOR TRAINING ARTIFICIAL NEURAL NETWORKS - A simplified explanation of the abstract
This abstract first appeared for US patent application 20240355320 titled 'SYSTEMS AND METHODS FOR TRAINING ARTIFICIAL NEURAL NETWORKS
The patent application describes a system that utilizes neural networks to enhance speech quality by training on noisy input data and generating clean speech predictions.
- The system includes one or more processors and memory with instructions for processing.
- It generates augmented input data by adding noise components to training data.
- A first neural network is trained on the augmented input data and ground truth data to predict clean speech.
- The trainable parameters of the first neural network are locked after training.
- A second neural network is then trained on the augmented input data and predictions of the first network to output a second prediction of clean speech.
Potential Applications: - Speech enhancement in communication devices - Noise reduction in audio recordings - Improving speech recognition systems
Problems Solved: - Enhancing speech quality in noisy environments - Improving the accuracy of speech recognition systems
Benefits: - Clearer and more accurate speech output - Enhanced user experience in communication devices - Increased efficiency in speech processing tasks
Commercial Applications: Title: "Advanced Speech Enhancement Technology for Communication Devices" This technology can be utilized in smartphones, smart speakers, and other communication devices to provide users with clearer and more accurate speech output, enhancing the overall user experience. The market implications include improved customer satisfaction and increased demand for devices with superior speech enhancement capabilities.
Questions about Speech Enhancement Technology: 1. How does this technology compare to traditional noise reduction methods? This technology surpasses traditional noise reduction methods by utilizing neural networks to learn and adapt to different noise environments, resulting in more accurate and effective speech enhancement.
2. Can this technology be integrated into existing speech recognition systems? Yes, this technology can be integrated into existing speech recognition systems to improve their accuracy and performance in noisy environments.
Original Abstract Submitted
a system including: one or more processors; and memory including instructions that, when executed by the one or more processors, cause the one or more processors to: generate augmented input data by mixing noise components of training data; train a first neural network based on the augmented input data and ground truth data of the training data to output a first prediction of clean speech; lock trainable parameters of the first neural network as a result of the training of the first neural network; and train a second neural network according to the augmented input data and predictions of the first neural network to output a second prediction of the clean speech.
