Program automatic monitoring system and method based on set top box transformation and AI analysis

Abstract

The invention provides an automatic program monitoring system and method based on set top box transformation and AI analysis, and relates to the technical field of digital culture, in the invention, a heterogeneous device is utilized to control a terminal simulation instruction, and an HDMI encoder is combined to realize stock asset streaming access and matrix polling; the transmission continuity of the weak network is guaranteed by adopting code rate adaptive logic, a parallel dual-core AI engine is constructed to fuse picture semantics, physical characteristics and MFCC audio fingerprints, and content violation and audio abnormity are accurately recognized; introducing a closed-loop self-healing strategy, judging a fault through a motion vector, and triggering a power supply to restart and reset; and constructing a weighted evaluation model based on the alarm severity, the operation and maintenance scene coefficient and the weight coefficient, and realizing adaptive distribution of response priorities. According to the method, the old utilization upgrading cost is effectively reduced, the deep fusion efficiency of technology monitoring and content supervision is improved through a privacy desensitization and model increment updating mechanism, and safe broadcasting of radio and television is guaranteed.

Description

Technical Field

[0001] This invention relates to the field of digital culture technology, specifically to an automatic program monitoring system and method based on set-top box modification and AI analysis. Background Technology

[0002] Automated program monitoring has evolved from early manual inspections to analog signal detection, and then to digital content recognition and intelligent semantic analysis. Early monitoring relied on manual labor or simple physical level detection, which was inefficient and unable to analyze content. With the rise of digital fingerprinting, watermarking, and multi-source heterogeneous acquisition technologies, monitoring methods have achieved a qualitative leap from "signal-level" to "content-level" recognition. Currently, through the deep integration of artificial intelligence technologies such as deep learning, computer vision (CV), and natural language processing (NLP), automated monitoring has evolved into an intelligent system with multi-dimensional semantic understanding and automated risk warning capabilities. It is widely used in fields such as broadcasting supervision, streaming media content review, advertising monitoring, and copyright protection, completing an intelligent leap from "post-event investigation" to "pre-event prevention and full coverage."

[0003] However, existing monitoring systems generally face the dilemma of high procurement costs for specialized equipment and the "isolation" of existing set-top box resources. Due to the lack of low-cost, non-intrusive automated transformation methods, it is difficult to achieve intelligent reuse and upgrading of massive existing assets. Furthermore, basic channel polling and scheduling work still heavily relies on inefficient, fatigue-prone, and missed detection manual operations.

[0004] Secondly, the existing system suffers from fragmented monitoring dimensions and rigid operation and maintenance models. Technical means are mostly focused on physical signal layer or basic image anomalies, lacking in-depth identification of content semantic compliance, resulting in a serious disconnect between technical monitoring and content supervision. At the same time, the system cannot adaptively adjust alarm distribution strategies according to different operation and maintenance scenarios such as "unattended" or "manned computer room", making it difficult to meet the requirements of modern broadcast security in terms of the accuracy and real-time performance of alarm response. Summary of the Invention

[0005] The purpose of this invention is to provide an automatic program monitoring system and method based on set-top box modification and AI analysis to solve the problems mentioned in the background art.

[0006] To achieve the above objectives, the present invention provides the following technical solution: An automatic program monitoring system and method based on set-top box modification and AI analysis, the specific steps of which include: S1. By connecting the control interface of the existing set-top box to the heterogeneous device control terminal, channel switching is simulated by human commands, and the baseband signal output by the set-top box is converted into a standard streaming media network stream using an HDMI encoder, so as to realize the cloud computing power access of the closed signal source. S2. According to the preset inspection strategy, the dispatch center directs multiple heterogeneous device control terminals to form a cluster and perform matrix channel rotation. At the same time, the signal encoding module senses changes in network bandwidth in real time and dynamically adjusts the video compression ratio and encoding format to ensure the continuity of the monitoring stream under different network environments. S3. Input the collected streaming media data into the video quality diagnosis engine and content semantic compliance engine that work in parallel, and simultaneously extract the physical quality features of the picture, video semantic features and audio fingerprint features. Based on multi-dimensional feature fusion calculation, identify picture abnormalities, content violations and illegal audio insertion status. S4. Based on the AI ​​analysis results, if the monitoring system determines that the fault is at the device level, it will trigger the relay through the hardware control module to perform a power-off and restart operation of the set-top box, thereby achieving self-healing of the fault. It will also dynamically select the alarm distribution path and response priority according to the logic preset of the current operation and maintenance mode.

[0007] Furthermore, the heterogeneous device control terminal uses a microcontroller to simulate infrared or serial control signals to achieve non-intrusive automated transformation of existing set-top boxes. The dispatch center controls each set-top box in the cluster to perform step-by-step or skip-step channel switching according to a set time step by issuing task instruction sets, so as to solve the technical defects of "islanding" of existing resources and low efficiency of manual inspection.

[0008] Furthermore, the signal encoding module has a bit rate adaptive control logic, which automatically switches between H.264 and H.265 encoding standards by monitoring the available network bandwidth in real time, and dynamically adjusts the quantization parameters to ensure that the packet loss rate of the monitored stream is lower than a preset threshold in a weak network environment.

[0009] Furthermore, the dual-core AI analysis engine includes a video quality analysis branch and a semantic compliance recognition branch that operate in parallel; the video quality analysis branch is used to identify still frames, black screens, color bars, and snow noise. The semantic compliance recognition branch integrates a deep learning model to compare sensitive figures, illegal texts, illegal logos, and sensitive scenes in the image in real time, so as to simultaneously perform technical monitoring and content supervision and deeply integrate them.

[0010] Furthermore, the dual-core AI analysis engine also integrates audio feature extraction and fingerprint comparison logic; Specifically, the acquired audio signals are digitally processed to extract Mel frequency cepstral coefficients used to characterize audio features; audio feature fingerprints are constructed based on Mel frequency cepstral coefficients; and the real-time extracted audio feature fingerprints are matched with a pre-stored legitimate program feature library for similarity. When the deviation between the real-time extracted audio feature fingerprint and the pre-stored legitimate program feature library exceeds a preset range, or when a specific noise energy distribution is detected, it is determined that the current program is in an abnormal state, including illegal audio insertion, mute, or noise.

[0011] Furthermore, step S4 includes a closed-loop self-healing strategy, specifically: when the AI ​​analysis module determines that the current channel has a continuous static frame or the signal interruption duration exceeds the set logic threshold, the system sends a power-off command to the external relay through the hardware control module, forcing the set-top box to restart and reset, thus completing the automatic repair of the physical link.

[0012] Furthermore, the content semantic compliance engine supports privacy desensitization processing strategies. By pre-configuring dynamic masking in the monitoring screen, pixel-level blocking or weight reduction of computing power allocation can be performed on specific areas such as advertising positions and specific logo positions to reduce computing power consumption in non-core areas and meet data compliance requirements.

[0013] Furthermore, the urgency of the alarm is determined using a multi-factor weighted evaluation model. The specific calculation process is as follows: The alarm severity factor is determined and automatically assigned by the AI ​​analysis engine based on the fault type, with sensitive content violations having a higher weight than technical black screens. An operation and maintenance scenario coefficient is introduced, which is dynamically adjusted according to the current status of the data center. In "unattended" mode, the operation and maintenance scenario coefficient will be increased to prevent missed detections. The alarm severity factor and the operation and maintenance scenario coefficient are multiplied by their respective weight coefficients and then summed, with a correction deviation constant added. The higher the comprehensive score, the more urgent the alarm, which triggers a higher-level distribution channel, including upgrading from ordinary APP push to mandatory telephone voice alarm.

[0014] Furthermore, the content semantic compliance engine has an AI model incremental update mechanism. By constructing a manual verification feedback loop, it collects labeled false or missed video frames and uses an active learning algorithm to fine-tune the online recognition model with small samples, thereby achieving continuous evolution of compliance recognition accuracy.

[0015] An automatic program monitoring system based on set-top box upgrades and AI analysis includes: The baseband streaming access module connects to the control interface of existing set-top boxes through a heterogeneous device control terminal, simulates manual commands to switch channels, and uses an HDMI encoder to convert the baseband signal output by the set-top box into a standard streaming media network stream, realizing cloud computing power access for closed signal sources. The cluster inspection and scheduling module, according to the preset inspection strategy, directs multiple heterogeneous device control terminals to form a cluster and perform matrix-style channel rotation. At the same time, the signal encoding module senses changes in network bandwidth in real time and dynamically adjusts the video compression ratio and encoding format to ensure the continuity of the monitoring stream under different network environments. The multi-dimensional intelligent analysis module inputs the collected streaming media data into the video quality diagnosis engine and content semantic compliance engine, which work in parallel. It simultaneously extracts the physical quality features of the picture, the semantic features of the video, and the audio fingerprint features. Based on the multi-dimensional feature fusion calculation, it identifies abnormal picture, content violation and illegal audio insertion status. The closed-loop self-healing alarm module, based on the AI ​​analysis results, if the monitoring system determines that it is a device-level fault, it will trigger a relay through the hardware control module to perform a power-off and restart operation of the set-top box, thereby achieving fault self-healing. It will also dynamically select the alarm distribution path and response priority according to the logic preset of the current operation and maintenance mode.

[0016] Compared with the prior art, the beneficial effects of the present invention are: This invention effectively solves the dilemma of high procurement costs for specialized equipment and the "isolation" of existing set-top box resources by working collaboratively with the baseband streaming access module and the cluster inspection and scheduling module. It utilizes non-intrusive heterogeneous device control terminals to simulate infrared or serial control signals, achieving low-cost intelligent upgrading of massive amounts of existing assets. This completely changes the previous situation where channel polling and scheduling relied heavily on manual operation, leading to low efficiency and easy fatigue-induced missed detections. The scheduling center's command and control module cluster performs matrix-style channel polling according to a set time step, greatly improving the automation level of large-sample monitoring. Simultaneously, in conjunction with the signal encoding module's adaptive bitrate control logic, it dynamically adjusts quantization parameters and automatically switches between H.264 and H.265 encoding standards by using real-time sensing of available network bandwidth as a decision variable. This ensures that the packet loss rate of the monitoring stream remains below a preset threshold even in weak network environments, achieving efficient and stable bridging of the closed signal source to cloud computing power.

[0017] This invention addresses the issues of fragmented monitoring dimensions and rigid operation and maintenance models through the deep integration of a multi-dimensional intelligent analysis module and a closed-loop self-healing alarm module. The system utilizes a parallel dual-core engine to simultaneously extract physical quality features, video semantic features, and audio fingerprint features based on Mel-frequency cepstral coefficients. It determines the state of still frames by calculating motion vectors between consecutive frames of the video stream and combines this with a fault self-healing success rate evaluation index, achieving a deep closed loop between technical monitoring and content supervision, effectively bridging the technological gap between the two. Furthermore, a multi-factor weighted evaluation model comprehensively considers the severity of alarms. The system calculates a comprehensive score reflecting the urgency of alarms by combining degree factors, operation and maintenance scenario coefficients, and their respective weight coefficients, and adding a correction deviation constant. This enables an adaptive alarm distribution strategy for different operation and maintenance scenarios such as "unattended" or "manned computer room". Combined with privacy desensitization logic that weights the spatial dimension of images using a dynamic masking matrix, and an AI model incremental update mechanism based on manual verification feedback loop, the system significantly improves the accuracy of alarm response and compliance identification accuracy while reducing computing power consumption in non-core areas, thus comprehensively meeting the real-time requirements of modern broadcast security. Attached Figure Description

[0018] Figure 1 This is a schematic diagram of the overall method flow of the present invention; Figure 2 This is a schematic diagram of the overall system framework structure in this invention. Detailed Implementation

[0019] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0020] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below. Example

[0021] Please see Figure 1 This invention provides a technical solution: an automatic program monitoring system and method based on set-top box modification and AI analysis, the specific steps of which include: S1. By connecting the control interface of the existing set-top box to the heterogeneous device control terminal, channel switching is simulated by human commands, and the baseband signal output by the set-top box is converted into a standard streaming media network stream using an HDMI encoder, so as to realize the cloud computing power access of the closed signal source. S2. According to the preset inspection strategy, the dispatch center directs multiple heterogeneous device control terminals to form a cluster and perform matrix channel rotation. At the same time, the signal encoding module senses changes in network bandwidth in real time and dynamically adjusts the video compression ratio and encoding format to ensure the continuity of the monitoring stream under different network environments. S3. Input the collected streaming media data into the video quality diagnosis engine and content semantic compliance engine that work in parallel, and simultaneously extract the physical quality features of the picture, video semantic features and audio fingerprint features. Based on multi-dimensional feature fusion calculation, identify picture abnormalities, content violations and illegal audio insertion status. S4. Based on the AI ​​analysis results, if the monitoring system determines that the fault is at the device level, it will trigger the relay through the hardware control module to perform a power-off and restart operation of the set-top box, thereby achieving self-healing of the fault. It will also dynamically select the alarm distribution path and response priority according to the logic preset of the current operation and maintenance mode.

[0022] In this embodiment, through the systematic connection of steps S1 to S4, a full-chain monitoring system from the reuse of existing hardware at the bottom layer to intelligent decision-making at the top layer is constructed, which significantly improves the automation level and operation and maintenance efficiency of broadcast monitoring services. Step S1 realizes the non-intrusive transformation of existing set-top boxes through heterogeneous device control terminals and HDMI encoders. Its significance lies in breaking the physical limitations of closed signal sources and realizing the cloud computing access of massive existing assets at extremely low cost. Step S2 utilizes the scheduling center to direct the cluster to perform matrix-style channel round-robin, and senses network bandwidth in real time to dynamically adjust quantization parameters and H.264 / H.265 encoding switching. Its significance lies in solving the pain point of low efficiency of manual round-robin inspection, and ensuring the transmission continuity of the monitoring stream in complex computer room environments and weak network conditions. Step S3 extracts physical features, semantic features, and audio fingerprint features based on Mel frequency cepstral coefficients from the image through a dual-core AI analysis engine. It also combines motion vectors to determine still frames and occlusion mask matrices to perform privacy desensitization. Its significance lies in achieving a deep integration of technical indicators and content supervision, and improving the accuracy of identifying illegal insertions, illegal content, and abnormal images through multi-dimensional feature calculation. Step S4, after determining a device-level fault, triggers a power-down restart logic via a relay. It also constructs a multi-factor weighted evaluation model by combining alarm severity factors, operation and maintenance scenario coefficients, and correction deviation constants. Its significance lies in achieving automated closed-loop self-healing of faults. Furthermore, through a hierarchical alarm distribution strategy, it ensures the system's second-level response and accurate handling of core faults under different operation and maintenance modes, such as "unattended operation."

[0023] The heterogeneous device control terminal uses a microcontroller to simulate infrared or serial control signals to achieve non-intrusive automated transformation of existing set-top boxes. The dispatch center controls each set-top box in the cluster to perform step-by-step or skip-step channel switching according to a set time step by issuing task instruction sets, so as to solve the technical defects of "islanding" of existing resources and low efficiency of manual inspection.

[0024] The signal encoding module has a code rate adaptive control logic. By monitoring the available network bandwidth in real time, it automatically switches between H.264 and H.265 encoding standards and dynamically adjusts the quantization parameters to ensure that the packet loss rate of the monitored stream is lower than a preset threshold in a weak network environment.

[0025] By sensing the current available network bandwidth in real time, the system uses this as a decision variable. When the bandwidth value fluctuates, the system automatically switches to a coding standard with higher encoding efficiency or better compatibility, i.e., switching between H.264 and H.265, and simultaneously adjusting the quantization parameters. The principle of this dynamic adjustment is to sacrifice some image quality clarity or increase the compression ratio in exchange for a lower transmission load, thereby ensuring that the packet loss rate of the monitored video stream remains below a safe threshold under bandwidth constraints, thus guaranteeing the continuity of the monitoring task.

[0026] The dual-core AI analysis engine includes a video quality analysis branch and a semantic compliance recognition branch that work in parallel; the video quality analysis branch is used to identify still frames, black screens, color bars, and snow noise. The semantic compliance recognition branch integrates a deep learning model to compare sensitive figures, illegal texts, illegal logos, and sensitive scenes in the image in real time, so as to simultaneously perform technical monitoring and content supervision and deeply integrate them.

[0027] The dual-core AI analysis engine also integrates audio feature extraction and fingerprint comparison logic; Specifically, the acquired audio signals are digitally processed to extract Mel frequency cepstral coefficients used to characterize audio features; audio feature fingerprints are constructed based on Mel frequency cepstral coefficients; and the real-time extracted audio feature fingerprints are matched with a pre-stored legitimate program feature library for similarity. When the deviation between the real-time extracted audio feature fingerprint and the pre-stored legitimate program feature library exceeds a preset range, or when a specific noise energy distribution is detected, it is determined that the current program is in an abnormal state, including illegal audio insertion, mute, or noise.

[0028] Step S4 includes a closed-loop self-healing strategy. The specific logic is as follows: when the AI ​​analysis module determines that the current channel has a continuous static frame or the signal interruption duration exceeds the set logic threshold, the system sends a power-off command to the external relay through the hardware control module to force the set-top box to restart and reset, thus completing the automatic repair of the physical link.

[0029] Furthermore, the logic for fault diagnosis and self-healing success rate assessment includes: Still Frame Determination: The AI ​​analysis module calculates the motion vectors between consecutive frames in the video stream. The logic is: if the pixel changes of multiple consecutive frames are extremely small, that is, the motion vector values ​​approach zero, the system determines that the current scene is frozen or static.

[0030] Self-healing performance evaluation: To quantify system performance, a fault self-healing success rate was defined. The calculation principle is to count the number of channels that successfully resumed normal playback after a "power-down restart" operation via a hardware module within a specific monitoring period, and then divide this number by the total number of faults detected by the system during that period. This ratio intuitively reflects the effectiveness of automated methods in replacing manual maintenance.

[0031] The content semantic compliance engine supports privacy desensitization processing strategies. By pre-configuring dynamic masking in the monitoring screen, pixel-level blocking or weight reduction of computing power allocation can be performed on specific areas such as advertising positions and specific logo positions to reduce computing power consumption in non-core areas and meet data compliance requirements.

[0032] The image masking logic specifically involves weighting the image matrix in a spatial dimension. The system predefines a dynamic masking matrix that covers areas of the image that do not require monitoring (including advertising corner marks). During AI calculations, the system reduces or sets the pixel weights of the masked areas to zero. This logic reduces feature extraction computation in non-core areas, focusing limited computing power on the main content of the program, significantly improving the system's frame rate while ensuring the accuracy of compliance review.

[0033] The urgency of alarms is determined using a multi-factor weighted evaluation model. The specific calculation process is as follows: The alarm severity factor is determined and automatically assigned by the AI ​​analysis engine based on the fault type, with sensitive content violations having a higher weight than technical black screens. An operation and maintenance scenario coefficient is introduced, which is dynamically adjusted according to the current status of the data center. In "unattended" mode, the operation and maintenance scenario coefficient will be increased to prevent missed detections. The alarm severity factor and the operation and maintenance scenario coefficient are multiplied by their respective weight coefficients and then summed, with a correction deviation constant added. The higher the comprehensive score, the more urgent the alarm, which triggers a higher-level distribution channel, including upgrading from ordinary APP push to mandatory telephone voice alarm.

[0034] The content semantic compliance engine has an AI model incremental update mechanism. By constructing a manual verification feedback loop, it collects labeled false or missed video frames and uses active learning algorithms to fine-tune the online recognition model with small samples, thereby achieving continuous evolution of compliance recognition accuracy.

[0035] In this embodiment, the heterogeneous device control terminal uses a microcontroller to simulate infrared or serial signals, driving the cluster to perform step-by-step or skip-step channel switching by setting a time step. This is significant in eliminating the "islanding" dilemma of existing resources and replacing inefficient manual operations. The signal encoding module uses the real-time perceived available network bandwidth as a decision variable, switching between H.264 and H.265 standards and dynamically adjusting quantization parameters. Its calculation logic is to sacrifice non-core image quality for transmission stability, ensuring that the packet loss rate of the monitored stream is always below a preset safety threshold. The dual-core AI analysis engine collects video pixel features and audio digital signals, using the logic that motion vectors between consecutive frames approach zero to determine still frame faults, and extracts Mel-frequency cepstral coefficients to form an audio fingerprint for similarity matching with a legal database. This is significant in achieving deep fusion identification of technical faults, content violations, and illegal insertions. The closed-loop self-healing strategy is implemented for signal interruption duration. When the logic threshold is exceeded, the relay is triggered to power down and restart. The self-healing success rate (i.e., the number of channels successfully restored divided by the total number of faults) is used as a quantitative indicator. Its evaluation meaning is to intuitively reflect the effectiveness of automated methods in repairing physical links. The content semantic compliance engine uses a predefined dynamic masking matrix to weight the image matrix, reducing or setting the weight of pixels in non-core areas to zero. Its calculation meaning is to focus computing power on the main content of the program to significantly improve the processing frame rate. Finally, the multi-factor weighted evaluation model calculates a comprehensive score by combining the alarm severity factor, operation and maintenance scenario coefficient, weight coefficient, and correction deviation constant. Its evaluation meaning is to accurately determine the urgency of the alarm and automatically trigger the hierarchical distribution channel from APP push to forced telephone voice. In addition, the model incremental update mechanism performs small sample fine-tuning by collecting labeled video frames. Its significance is to realize the continuous evolution of the system's recognition accuracy driven by human feedback.

[0036] Please see Figure 2 An automatic program monitoring system based on set-top box modification and AI analysis includes: The baseband streaming access module connects to the control interface of existing set-top boxes through a heterogeneous device control terminal, simulates manual commands to switch channels, and uses an HDMI encoder to convert the baseband signal output by the set-top box into a standard streaming media network stream, realizing cloud computing power access for closed signal sources. The cluster inspection and scheduling module, according to the preset inspection strategy, directs multiple heterogeneous device control terminals to form a cluster and perform matrix-style channel rotation. At the same time, the signal encoding module senses changes in network bandwidth in real time and dynamically adjusts the video compression ratio and encoding format to ensure the continuity of the monitoring stream under different network environments. The multi-dimensional intelligent analysis module inputs the collected streaming media data into the video quality diagnosis engine and content semantic compliance engine, which work in parallel. It simultaneously extracts the physical quality features of the picture, the semantic features of the video, and the audio fingerprint features. Based on the multi-dimensional feature fusion calculation, it identifies abnormal picture, content violation and illegal audio insertion status. The closed-loop self-healing alarm module, based on the AI ​​analysis results, if the monitoring system determines that it is a device-level fault, it will trigger a relay through the hardware control module to perform a power-off and restart operation of the set-top box, thereby achieving fault self-healing. It will also dynamically select the alarm distribution path and response priority according to the logic preset of the current operation and maintenance mode.

[0037] It should be noted that all calculation formulas in this application employ regression analysis, including but not limited to machine learning algorithms, to deeply analyze the collected parameters and identify their natural trends and interrelationships. Specialized software, such as Python's Scikit-learn library or the R language, is used to automatically generate mathematical models that match the data. Then, cross-validation and other methods are used to objectively evaluate the model performance, and continuous feedback and optimization are combined to ensure that the created formulas truly reflect the inherent laws of the data, thereby guaranteeing their effectiveness and accuracy. In all calculation formulas in this application, the parameters in each formula undergo dimensionless processing within a consistent range to ensure that different physical quantities are compared on the same scale; dimensionless processing techniques include, but are not limited to, min-max-normalization and Z-score standardization. The technical solution of this invention, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as a computer floppy disk, read-only memory (ROM), random-access memory (RAM), flash memory, hard disk, or optical disk, etc., including several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods of various embodiments of this invention.

[0038] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as a sequenced list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (such as a computer-based system, a processor-including system, or other system that can fetch and execute instructions from, an instruction execution system, apparatus, or device). For the purposes of this specification, "computer-readable medium" can be any means that can contain, store, communicate, propagate, or transmit programs for use by, or in conjunction with, an instruction execution system, apparatus, or device.

[0039] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. An automatic program monitoring system and method based on set-top box modification and AI analysis, characterized in that, The specific steps include: S1. By connecting the control interface of the existing set-top box to the heterogeneous device control terminal, channel switching is simulated by human commands, and the baseband signal output by the set-top box is converted into a standard streaming media network stream using an HDMI encoder, so as to realize the cloud computing power access of the closed signal source. S2. According to the preset inspection strategy, the dispatch center directs multiple heterogeneous device control terminals to form a cluster and perform matrix channel rotation. At the same time, the signal encoding module senses changes in network bandwidth in real time and dynamically adjusts the video compression ratio and encoding format to ensure the continuity of the monitoring stream under different network environments. S3. Input the collected streaming media data into the video quality diagnosis engine and content semantic compliance engine that work in parallel, and simultaneously extract the physical quality features of the picture, video semantic features and audio fingerprint features. Based on multi-dimensional feature fusion calculation, identify picture abnormalities, content violations and illegal audio insertion status. S4. Based on the AI ​​analysis results, if the monitoring system determines that the fault is at the device level, it will trigger the relay through the hardware control module to perform a power-off and restart operation of the set-top box, thereby achieving self-healing of the fault. It will also dynamically select the alarm distribution path and response priority according to the logic preset of the current operation and maintenance mode.

2. The automatic program monitoring method based on set-top box modification and AI analysis according to claim 1, characterized in that: The heterogeneous device control terminal uses a microcontroller to simulate infrared or serial control signals to achieve non-intrusive automated transformation of existing set-top boxes. The dispatch center controls each set-top box in the cluster to perform step-by-step or skip-step channel switching according to a set time step by issuing task instruction sets, so as to solve the technical defects of "islanding" of existing resources and low efficiency of manual inspection.

3. The automatic program monitoring method based on set-top box modification and AI analysis according to claim 2, characterized in that: The signal encoding module has a code rate adaptive control logic. By monitoring the available network bandwidth in real time, it automatically switches between H.264 and H.265 encoding standards and dynamically adjusts the quantization parameters to ensure that the packet loss rate of the monitored stream is lower than a preset threshold in a weak network environment.

4. The automatic program monitoring method based on set-top box modification and AI analysis according to claim 3, characterized in that: The dual-core AI analysis engine includes a video quality analysis branch and a semantic compliance recognition branch that work in parallel; the video quality analysis branch is used to identify still frames, black screens, color bars, and snow noise. The semantic compliance recognition branch integrates a deep learning model to compare sensitive figures, illegal texts, illegal logos, and sensitive scenes in the image in real time, so as to simultaneously perform technical monitoring and content supervision and deeply integrate them.

5. The automatic program monitoring method based on set-top box modification and AI analysis according to claim 4, characterized in that: The dual-core AI analysis engine also integrates audio feature extraction and fingerprint comparison logic; Specifically, the acquired audio signals are digitally processed to extract Mel-frequency cepstral coefficients used to characterize audio features; and audio feature fingerprints are constructed based on the Mel-frequency cepstral coefficients. The real-time extracted audio feature fingerprints are matched with a pre-stored database of legitimate program features for similarity. When the deviation between the real-time extracted audio feature fingerprint and the pre-stored legitimate program feature library exceeds a preset range, or when a specific noise energy distribution is detected, it is determined that the current program is in an abnormal state, including illegal audio insertion, mute, or noise.

6. The automatic program monitoring method based on set-top box modification and AI analysis according to claim 5, characterized in that: Step S4 includes a closed-loop self-healing strategy. The specific logic is as follows: when the AI ​​analysis module determines that the current channel has a continuous static frame or the signal interruption duration exceeds the set logic threshold, the system sends a power-off command to the external relay through the hardware control module to force the set-top box to restart and reset, thus completing the automatic repair of the physical link.

7. The automatic program monitoring method based on set-top box modification and AI analysis according to claim 6, characterized in that: The content semantic compliance engine supports privacy desensitization processing strategies. By pre-configuring dynamic masking in the monitoring screen, pixel-level blocking or weight reduction of computing power allocation can be performed on specific areas such as advertising positions and specific logo positions to reduce computing power consumption in non-core areas and meet data compliance requirements.

8. The automatic program monitoring method based on set-top box modification and AI analysis according to claim 7, characterized in that: The urgency of alarms is determined using a multi-factor weighted evaluation model. The specific calculation process is as follows: The alarm severity factor is determined and automatically assigned by the AI ​​analysis engine based on the fault type, with sensitive content violations having a higher weight than technical black screens. An operation and maintenance scenario coefficient is introduced, which is dynamically adjusted according to the current status of the data center. In "unattended" mode, the operation and maintenance scenario coefficient will be increased to prevent missed detections. The alarm severity factor and the operation and maintenance scenario coefficient are multiplied by their respective weight coefficients and then summed, with a correction deviation constant added. The higher the comprehensive score, the more urgent the alarm, which triggers a higher-level distribution channel, including upgrading from ordinary APP push to mandatory telephone voice alarm.

9. The automatic program monitoring method based on set-top box modification and AI analysis according to claim 8, characterized in that: The content semantic compliance engine has an AI model incremental update mechanism. By constructing a manual verification feedback loop, it collects labeled false or missed video frames and uses active learning algorithms to fine-tune the online recognition model with small samples, thereby achieving continuous evolution of compliance recognition accuracy.

10. An automatic program monitoring system based on set-top box modification and AI analysis, comprising the automatic program monitoring method based on set-top box modification and AI analysis as described in any one of claims 1 to 9, characterized in that, include: The baseband streaming access module connects to the control interface of existing set-top boxes through a heterogeneous device control terminal, simulates manual commands to switch channels, and uses an HDMI encoder to convert the baseband signal output by the set-top box into a standard streaming media network stream, realizing cloud computing power access for closed signal sources. The cluster inspection and scheduling module, according to the preset inspection strategy, directs multiple heterogeneous device control terminals to form a cluster and perform matrix-style channel rotation. At the same time, the signal encoding module senses changes in network bandwidth in real time and dynamically adjusts the video compression ratio and encoding format to ensure the continuity of the monitoring stream under different network environments. The multi-dimensional intelligent analysis module inputs the collected streaming media data into the video quality diagnosis engine and content semantic compliance engine, which work in parallel. It simultaneously extracts the physical quality features of the picture, the semantic features of the video, and the audio fingerprint features. Based on the multi-dimensional feature fusion calculation, it identifies abnormal picture, content violation and illegal audio insertion status. The closed-loop self-healing alarm module, based on the AI ​​analysis results, if the monitoring system determines that it is a device-level fault, it will trigger a relay through the hardware control module to perform a power-off and restart operation of the set-top box, thereby achieving fault self-healing. It will also dynamically select the alarm distribution path and response priority according to the logic preset of the current operation and maintenance mode.

Images