The application relates to the fields of
signal processing and intelligent detection, and specifically discloses a
HDMI signal intelligent detection and real-time
analysis method and
system. The method comprises the following steps: based on image frames, audio frames and control signaling frame sequences, key segments are extracted and compressed and encoded,
key frame indexes and original feature marks are generated; based on jump and fluctuation characteristics, abnormal indexes and candidate error segments are generated; multi-
modal fusion analysis is performed, diagnosis results and
modal feature vectors are generated; based on
modal features and device historical data, a device portrait is constructed, and detection parameter strategies are extracted; according to this, strategy optimization and
configuration analysis are performed, control information is generated, and light
model parameters are generated in combination with device features; finally,
edge model parameter synchronization and detection behavior closed-
loop optimization are completed. Through the application, high-precision real-time detection under
resource constraints can be realized on a differentiated
HDMI access device, the model deployment flexibility and detection response efficiency are effectively improved, and the application has strong
system adaptability and
engineering practical value.