17421073. METHOD AND APPARATUS FOR ACTION RECOGNITION simplified abstract (SAMSUNG ELECTRONICS CO., LTD.)
Contents
METHOD AND APPARATUS FOR ACTION RECOGNITION
Organization Name
Inventor(s)
Brais Martinez of Staines (GB)
Victor Augusto Escorcia of Staines (GB)
Juan Perez-rua of Staines (GB)
Antoine Toisoul of Staines (GB)
METHOD AND APPARATUS FOR ACTION RECOGNITION - A simplified explanation of the abstract
This abstract first appeared for US patent application 17421073 titled 'METHOD AND APPARATUS FOR ACTION RECOGNITION
Simplified Explanation
The present techniques involve a method and apparatus for action recognition, specifically on lightweight devices like smartphones. The aim is to adjust the machine learning (ML) model to achieve the desired accuracy and efficiency levels while considering the computational capability of the device.
- Adjusting the number of channels assigned to the full temporal resolution channels in the ML model.
- Adjusting the point in the ML model where the temporal pooling layer or layers are applied.
Potential Applications
This technology can be applied in various fields where action recognition is required, such as:
- Mobile applications for fitness tracking and exercise monitoring.
- Augmented reality and virtual reality applications for gesture recognition.
- Security systems for detecting suspicious or abnormal actions.
- Robotics for understanding and responding to human actions.
Problems Solved
The technology addresses the following problems:
- Limited computational capability of lightweight devices like smartphones.
- Balancing the trade-off between accuracy and efficiency in action recognition.
- Adapting the ML model to resource-constrained devices without sacrificing performance.
Benefits
The benefits of this technology include:
- Improved action recognition on resource-constrained devices.
- Optimal utilization of computational resources.
- Customizable ML models for different devices and applications.
- Enhanced user experience in various domains, including fitness, gaming, and security.
Original Abstract Submitted
Broadly speaking, the present techniques relate to a method and apparatus for performing action recognition, and in particular to a computer-implemented method for performing action recognition on resource-constrained or lightweight devices such as smartphones. The ML model may be adjusted to achieve required accuracy and efficiency levels, while also taking into account the computational capability of the apparatus that is being used to implement the ML model. One way is to adjust the number of channels assigned to the first set of channels, i.e. the full temporal resolution channels. Another way is to adjust the point in the ML model where the temporal pooling layer or layers are applied.
