Adaptive periodic signal monitoring system and edge device

Abstract

An adaptive periodic signal monitoring system includes a cloud server and an edge device. The edge device includes a communication module, a storage module, a sound receiving module, and an edge processing module. When the edge processing module determines that sound detection data from the sound receiving module contains a regular signal, it transmits the regular signal to the cloud server. The cloud server then uses an artificial intelligence model to obtain a feature vector group representing normal behavior, thereby obtaining multiple monitoring rules. The edge processing module stores these monitoring rules in the storage module, and a first converter module uses these monitoring rules to judge newly detected regular signals. If the converted regular signal is determined to be abnormal, an abnormal event signal is generated.

Claims

1. An adaptive periodic signal monitoring system, applicable to a field, comprising: a cloud server; and an edge device including a communication module for providing networking functionality, a storage module for data storage, and storing a first converter module, the first converter module serving as a data adapter to efficiently convert real-time signals into a format suitable for rule comparison and output comparison results; A microphone module continuously monitors the sound in the area to generate sound detection data, and an edge processing module is electrically connected to the communication module, the storage module, and the microphone module. The edge processing module receives the sound detection data from the microphone module and, when it determines that the sound detection data contains a regular signal, transmits the regular signal to the cloud server via the communication module. The cloud server stores the regular signal and processes it within a predetermined time interval using an artificial intelligence model to obtain a feature vector group representing normal behavior. A second converter module then transforms this feature vector group to obtain multiple monitoring rules, which are then transmitted back to the edge device. The artificial intelligence model includes an Audio Compact Deep Network (ACDNet) and a Graph Neural Network (GNN). The second converter module acts as a rule generator to translate high-dimensional feature vectors into specific rules that can be referenced by low-power edge devices. The edge processing module stores these monitoring rules in the storage module and extracts multiple features from the regular signals contained in the newly detected sound detection data by the audio module through a feature extractor. These features are then input into the first converter module, which uses these monitoring rules to compare with the features of the newly detected regular signals and outputs a comparison result. When the newly detected regular signals are determined to be abnormal based on the comparison result, an abnormal event signal is generated.

2. The adaptive periodic signal monitoring system as described in claim 1, wherein, The artificial intelligence model also includes a spiking neural network (SNN).

3. The adaptive periodic signal monitoring system as described in claim 1, wherein, The edge processing module first cuts the regular signal into a standardized form with a fixed length or pattern before transmitting it to the cloud server via the communication module.

4. The adaptive periodic signal monitoring system as described in claim 1, wherein, After obtaining the monitoring rules, the edge processing module continues to transmit the newly detected regular signals to the cloud server via the communication module. This allows the cloud server to periodically process the newly added regular signals using the artificial intelligence model to obtain new feature vector groups. When a significant statistical variation is detected between the original feature vector group and the new feature vector group, a statistical warning signal is generated.

5. The adaptive periodic signal monitoring system as described in claim 1, wherein, The cloud server also uses the artificial intelligence model to process the regular signals within each predetermined time interval to obtain the corresponding feature vector group, which becomes a feature vector historical information. A generative artificial intelligence model then analyzes the trend of the signal from the feature vector historical information.

6. An edge device suitable for a field and a cloud server, comprising: a communication module for providing networking functionality; a storage module for data storage, and storing a first converter module that acts as a data adapter to efficiently convert real-time signals into a format suitable for rule comparison and output comparison results; A microphone module continuously monitors the sound in the area to generate sound detection data; and an edge processing module electrically connected to the communication module, the storage module, and the microphone module, receives the sound detection data from the microphone module, and when it determines that the sound detection data contains a regular signal, transmits the regular signal to the cloud server via the communication module, so that the cloud server stores the regular signal, processes the regular signal within a predetermined time interval using an artificial intelligence model to obtain a feature vector group representing normal behavior, and transforms the feature vector group using a second converter module to obtain multiple monitoring rules, and transmits these monitoring rules back to the edge device. The artificial intelligence model includes an Audio Compact Deep Network (ACDNet) and a Graph Neural Network (GNN). The edge processing module stores the monitoring rules in the storage module and extracts multiple features from the regular signals contained in the newly detected sound detection data by the sound receiving module through a feature extractor. These features are then input into the first converter module, which compares the monitoring rules with the features of the newly detected regular signals and outputs a comparison result. When the edge processing module determines that the newly detected regular signals are abnormal based on the comparison result, it generates an abnormal event signal.

7. The edge device as described in claim 6, wherein, The edge processing module first cuts the regular signal into a standardized form with a fixed length or pattern before transmitting it to the cloud server via the communication module.

8. The edge device as described in claim 6, wherein, After obtaining the monitoring rules, the edge processing module continues to transmit the newly detected regular signals to the cloud server via the communication module. This allows the cloud server to periodically process the newly added regular signals using the artificial intelligence model to obtain new feature vector groups. When a significant statistical variation is detected between the original feature vector group and the new feature vector group, a statistical warning signal is generated.

Images