AI Stream Position Detection Using VPN and Audio-Video Matching
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing server systems struggle to accurately detect specific positions within data streams, such as commercial breaks or product placements, in video and audio content, especially when they do not match the predefined criteria of user devices due to regional differences.
Innovation Solution
A method and system that utilizes a client-server system with a Virtual Private Network (VPN) connection to receive and analyze data streams, combining AI-based video and audio analysis to enhance detection accuracy by creating screenshots and audio fingerprints, synchronized with timestamp analysis.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a server system uses AI-based video and audio analysis to detect specific positions, then detection accuracy is improved, but device complexity increases
Solution Approach 1:
The system segments the detection task into separate video analysis and audio analysis components, each processed independently before combining results. The video stream is processed to generate screenshots and the audio stream is processed to generate fingerprints, which are then compared to detect specific positions. This segmentation allows each component to be optimized independently while reducing overall system complexity.
Solution Approach 2:
The server system is designed to handle multiple data streams from different sources and apply the same AI-based detection methodology across various content types. The system can process different video and audio streams using universal AI models, making the complex detection system adaptable to multiple scenarios without requiring separate specialized systems for each content type.
2Adaptability or versatility
If the server system connects through a client-server system with VPN to access region-specific data streams, then adaptability to regional criteria is improved, but device complexity increases
Solution Approach 1:
The patent introduces a client-server system with VPN as an intermediary between the server and the region-specific data stream source. This intermediary enables the server to access region-specific content by routing connections through the VPN network, allowing the server to adapt to regional criteria without directly managing complex regional infrastructure. The VPN acts as a mediator that simplifies the connection architecture while providing the necessary regional access.
3Reliability
If the system processes both video and audio streams simultaneously for detection, then reliability of position detection is improved, but use of energy increases
Solution Approach 1:
The system performs preliminary processing of both video and audio streams independently before combining the results for final detection. Preprocessing tasks include generating screenshots from video frames and creating audio fingerprints, which can be done in parallel. This preliminary action allows the system to prepare data structures that facilitate faster comparison and reduce the computational burden during the actual detection phase, thereby managing energy consumption while maintaining high reliability.
Data Source
AI summary
A method features receiving a data stream (DS) having video and audio data; detecting a specific position within the DS classified by a predefined criterion (PC) and received by user's clients matching the PC; connecting the receiver not matching the PC with a client-server system matching the PC; receiving the DS by the client-server system and forwarding it to the receiver; extracting the video data and the audio data from the DS stream; creating a screenshot out of the video data and an associated timestamp; feeding the screenshot to an AI-system trained to detect whether it belongs to a predefined category; generating audio data corresponding with the screenshot and analyzing it for detecting whether it belongs to the predefined category; and if both the screenshot and the audio data generated/analyzed belong to the predefined category, then deciding that the associated timestamp for the screenshot defines the specific position.