Design And Implementation Of An Airborne Speech Signal Control System

In the control of aircraft,using speech to express the pilot's control operation,can be controlled more convenient intentions,especially in race against time of many operating combinations.But the existence of high intensity noise in the cabin,the selection of airborne speech control instructions and different accent of pilot,it brought difficulties to the application of speech control,running in embedded platform is one of factors influencing the speech control applications.So the paper designed a airborne speech signal control system using in the cabin,used to study the application of speech control technology challenges and difficulties.The airborne speech signal control system chooses ADSP-BF561 of the ADI company as the system hardware platform,realizing the speech input through the microphone on the pilot masks,collecting and processing speech by AD1939,and DMA controller is responsible for the transmission from speech signal to data buffer.Getting the speech recognition results through callingg the isolated word speech recognition algorithm,on the one hand,the results will transmit to the pilot headphones,on the other hand the results transmit to mission system by GJB289 A bus.In addition,the W5200 chip realizes the loading of speech model library,the testing of speech recognition performance and software upgrade.The paper uses a VAD algorithm based on Mel frequency cepstrum coefficients distance with an order statistics filter and pitch features,compared with the traditional method based on energy and the spectrum characteristics,it can get ideal results in noise environment,in the low SNR is still effective,and eliminate breathing interference under the mask.Using the key of cockpit as control instruction,can guarantee the consistency of pilot phonetic pronunciation and visual consistency,can be avoid unnecessary confusion.Training speech model through speaker adaptive technology can improve speech recognition rate of individual pilots.ADSP-BF561 is a fixed-point and dual-core processor,so the algorithm of speech recognition need to be fixed point and optimized identification process for the real-time control.The details of paper are as follows:(1)This paper analysed problems,including the existing airborne background noise,airborne voice commands,and the phonetic difference technology,presented system design scheme in airborne platform,and described the system design of hardware and software requirements.(2)The paper presented three optimization methods for improving the speech recognition performance:selection of acoustic model,optimization of endpoint detection algorithm and optimization of recognition program,in the end,designed the testing of speech recognition performance,the loading of speech model and speech interaction process.(3)Design software architecture of embedded system,and describes the realization of software process in flow chart.Under the simulation environment,the testing system,test results show that each instruction speech recognition rate achieves about 98%,All that serves to prove the practical use of this paper.