Research And Implementation Of Emergency Broadcasting System And Speech Enhancement

Under the influence of global warming,China's sudden natural disasters are more frequent,and due to the acceleration of China's urbanization,the loss of life and property caused by public safety emergencies will be further expanded.The grief brought about by the Wenchuan earthquake and the southern snow disaster in2008 is still vivid in our mind.It is urgent for our country to set up a complete set of emergency broadcasting system with safety,stability and timeliness.It is urgent for our country to set up a complete emergency broadcasting system with safety,stability and timeliness.When the disaster comes,it will guide the victims to take refuge in time and reduce the loss of life and property to the lowest.At the same time,speech signal,as the main carrier of information transmission in the emergency broadcasting system,is easily disturbed by noise during transmission or acquisition,which greatly reduces the effect of the emergency broadcast system.Therefore,it is of great significance to study the speech enhancement method in emergency broadcast system.This thesis aims to provide a stable,reliable and secure emergency broadcasting system solution,and studies the construction of the system,hardware and software collaboration,voice enhancement,decision making mechanism.The main research work of this thesis is as follows:(1)A multi-band spectral subtraction speech enhancement algorithm based on Mel domain and a real-time endpoint detection method based on noise characteristics,are proposed.According to the demand of timeliness in emergency broadcasting system,the subjective evaluation results of different speech enhancement algorithms are analyzed,and the multi band spectral subtraction,which has low complexity and strong robustness under different noise environment,is introduced into the emergency broadcast system for speech enhancement.However,due to the poor ability of the traditional multi-band subtraction method to distinguish the intensity of noise interference,on the basis of human auditory perception mechanism,a multi-band subtraction method based on the Mel domain banding method is proposed in this paper.The double-threshold method for endpoint detection cannot be introduced into the real-time system,and a real-time endpoint detection method is proposed by combining the entropy ratio calculation factor and the noise characteristics.(2)Construct an emergency broadcasting system optimization model with multi-level linkage,rapid response,extensive deployment and peacetime and wartime integration.According to the uncertainties of emergency broadcasting operation environment and the elements,behavior and rules of each level emergency broadcast system,the optimization model of emergency broadcasting system is proposed based on the ideas of multi-level cooperative management,event driven and recovery design.(3)Design and implement an emergency broadcast system solution based on Zynq7020 heterogeneous architecture chip.Based on the proposed optimization model of emergency broadcasting system,considering the peace-and-warfare mode switching,multi-level concurrent preemption design source scheduling decision mechanism,using the FPGA parallel operation characteristics and the AXI high speed communication principle,the high concurrency receiving processing module can be realized to achieve high stability and fast response.