Research On Sound Recognition Algorithm Based On Multi-Channel Temporal-Spectral Attention Model And Design Of Cable Trench External Force Damage Early Warning System

Nowadays,in the process urban planning and construction,most transmission lines are laid under the ground.However,underground cables laid in cable ditches are easily affected by external factors such as road excavation and pipeline burial.These construction processes may produce different degrees of damage to the cable,which in turn seriously affects the safety and reliability of power supply,causing huge losses to the masses in the power-using areas as well as to the power grid units.At present,in order to prevent the occurrence of such accidents,the power grid units strengthen the site inspection and inspection of key transmission lines and prohibit the work of engineering equipment around the cable ditch.However,this treatment method requires high human and material resources costs,so it is necessary to design an automated early warning system,ensuring timely warning before cables are damaged by external forces and realizing round-the-clock protection of underground cable lines cable lines.In this paper,the following four aspects of research work are carried out for cable ditch external force damage early warning system:(1)To take acoustic propagation characteristics of breaking machine for cable ditch as the research object,analyzing the relationship between sound signal propagation attenuation rate and influencing factors including distance and sound signal frequency,and analyzing the impact of interference signals in the process of sound propagation to determine the cable ditch sound sensor installation location and the cable ditch safety distance threshold.(2)Based on the characteristics of the sound signals in external damage events to cable ditches,the feature expression ability of different features for external damage sound events is analyzed.A data enhancement method is used to solve the overfitting problem of model training caused by the insufficient number and diversity of external breakage sound samples,so that the recognition performance of the model is more stable when dealing with changes in sound data.(3)Based on the characteristics of the external damage sound recognition task,an external damage sound recognition model is proposed.The model simultaneously learns the distinction between external damage sound events and security events as well as the feature differences between individual external damage sound events to reason about the possibility of the existence of damage events and the specific types of damage events.In addition,to reduce the influence of background noise on the recognition performance of the model,attentional feature enhancement modules such as temporal attention,frequency attention and channel attention are used to enhance or suppress sound features,and a parallel multi-channel time-frequency attention mechanism is used to improve the computational efficiency of the model.(4)Based on the analysis of the functional requirements and performance requirements of the system,the cable ditch external force damage early warning system was designed and implemented,and the effectiveness of the system’s warning function for external damage events was verified during the test application of actual scenarios.