Research On Speech Recognition Technology For Unmanned Combat Platform Control

Unmanned combat platforms have significant advantages in the fields of anti-terrorism and anti-riot,military reconnaissance,and rapid combat on the battlefield.They can enable soldiers to stay away from harsh battlefields and greatly retain the vitality of the army.Most of the traditional control methods of unmanned combat platforms require manual terminal control.It is difficult to free the operator's hands and eyes from the control handle and terminal screen,which limits the operator's combat effectiveness.To this end,this paper takes voice recognition and control technology as a new control method for unmanned combat platforms,and conducts research on voice recognition system design and control of unmanned combat platforms.First,the research on speech recognition algorithms in the battlefield environment is carried out.The typical gunshots are selected as the battlefield noise background,the noise characteristics of the battle environment are studied.The time-domain waveforms and spectrograms of the speech under the gunshot background are analyzed,and its locally stable noise characteristics are obtained.Six classical noise reduction algorithms are compared and analyzed through experiments,and the signal-to-noise ratio gain,subjective quality evaluation,running time and other indicators of each noise reduction technology are comprehensively considered,and finally the Least Mean Square adaptive filtering algorithm is determined as the optimal noise reduction method.Secondly,the Hidden Markov Model based on Gaussian Mixture Model is selected as the speech pattern matching algorithm,and the Gaussian Mixture Model is selected as the speaker recognition algorithm,and the speaker recognition system is built.Through experiments,the influence of each model parameter on the recognition result is explored,and appropriate speech model parameters are determined,which verified the performance of the recognition system in a noisy environment.In a noise environment of-10 d B,the speech recognition accuracy rate remains above 80%,and the speaker recognition accuracy rate is above 84%.Finally,a test prototype of a voice-controlled unmanned platform is built on an unmanned tracked vehicle based on the Arduino system,which can complete command actions through the Wi Fi remote voice control platform.The test verifies the feasibility and reliability of the voice control method.Real-time voice recognition is performed in an outdoor environment,and the accuracy of command recognition and speaker recognition are maintained above 90%.