Action Recognition Device Key Point Candidate Filtering
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing bottom-up action recognition schemes generate repeated or erroneous key point connection candidates, leading to decreased accuracy in action recognition results.
Innovation Solution
An action recognition device and method that includes a key point recognition unit, a key point connection unit, a connection candidate generating unit, and a connection candidate determining unit to select valid key point connection candidates, improving the accuracy of action recognition by filtering out redundant or erroneous candidates.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If bottom-up action recognition scheme is used to detect key points and generate connection candidates, then action recognition can be performed without pre-defining body parts, but repeated or erroneous key point connection candidates are generated leading to decreased accuracy
Solution Approach 1:
The patent extracts and removes repeated or erroneous key point connection candidates from the set of candidates before action recognition. By identifying and taking out invalid candidates through comparison of connection characteristics (such as spatial relationships between key points), the system maintains the flexibility of bottom-up approach while improving accuracy by eliminating harmful elements from the candidate pool.
Solution Approach 2:
The patent introduces a feedback mechanism where the system compares and evaluates key point connection candidates based on their geometric and spatial characteristics. This feedback loop identifies repeated or erroneous connections and uses this information to select valid candidates for action recognition, thereby improving measurement precision while preserving the adaptive nature of bottom-up processing.
2Productivity
If multiple key point connection candidates are generated without selection, then comprehensive coverage of possible connections is achieved, but processing time increases and accuracy decreases due to redundant candidates
Solution Approach 1:
The patent extracts and removes redundant key point connection candidates from the complete set of possible connections. By identifying and taking out repeated or erroneous candidates through systematic comparison of connection characteristics, the system maintains comprehensive coverage of valid connections while significantly reducing processing time by eliminating unnecessary redundant elements.
Solution Approach 2:
The patent discards repeated or erroneous key point connection candidates that do not contribute to accurate action recognition. By selectively discarding invalid candidates and recovering only the essential valid connections for further processing, the system achieves both comprehensive coverage of meaningful connections and efficient processing time.
Data Source
AI summary
Embodiments of this disclosure provide an action recognition device and method and an electronic device. The method includes: performing key point recognition on an object in a video frame by using a neural network to Obtain key point information and part affinity field score(s) of the object; performing key point connection according to the key point information and the part affinity field score(s); generating multiple key point connection candidates according to a result of the key point connection; for at least two of the multiple key point connection candidates, determining whether one of the at least two key point connection candidates is valid, so as to perform selection on the multiple key point connection candidates; and performing action recognition on the object according to the selected key point connection candidates. Hence, by selecting the generated key point connection candidates, accurate of action recognition in a bottom-up scheme may be improved.