Voice comparison and selection method for very high frequency voice communication system

Abstract

The invention discloses a voice comparison and selection method for a very high frequency voice communication system, relates to the field of very high frequency communication, and solves the problems that the signal-to-noise ratio calculation is not accurate enough, the delay is large, and the practicability is not strong enough in the comparison and selection processing process. The method comprises the following steps: carrying out first rough judgment on a voice normalization autocorrelation function envelope, then carrying out second judgment by taking a short-time average zero-crossing rate as a criterion, and screening out voice segments of multiple paths of signals; performing synchronization processing on the voice segments of the multipath signals, and selecting a non-voice segment from the voice synchronization signal after normalization processing as a noise segment; and subtracting the noise segment from the voice segment of the multiple paths of signals of voice synchronization after normalization processing to obtain a signal which is defaulted to be pure voice, and selecting one path of signal with the best signal-to-noise ratio as a comparison and selection result. According to the invention, the accuracy of voice comparison and selection and the environment adaptive capability are improved, and the processing delay of voice comparison and selection is shortened.

Description

technical field [0001] The invention relates to VHF communication, in particular to a voice comparison and selection method for a VHF voice communication system. Background technique [0002] At present, civil aviation VHF communication is still dominated by voice communication. Most of the air traffic control areas are large in size, and a single station cannot fully cover them. Generally, multiple stations are needed to achieve full coverage of the control sector, because multiple stations The locations are different, so the received signal strength of the same aircraft is different, and the interference intensity is also different. Naturally, the voice quality obtained by demodulation is also different. You can choose the best way through voice comparison. The voice signal is sent to the control center for the controller to monitor. [0003] The traditional voice comparison method is to directly compare the signal-to-noise ratio of multiple voice signals, and judge and s...

AI Technical Summary

Problems solved by technology

Because speech is divided into accent and unvoiced, in the same noise environment, the speech energy of accent and unvoiced is different, obviously the energy of accent is greater, the value of the calculated signal-to-noise ratio will be higher, so the situation of speech out-of-sync through the signal Noise ratio for voice comparison selection, the results obtained are not very reliable

[0004] There are three traditional voice signal-to-noise ratio calculation methods. The first one is out-of-band noise calculation. According to the spectrum characteristics of the voice signal, the voice signal is distributed within 3.4kHz, and the signal outside 3.4kHz is out-of-band noise. The voice signal-to-noise ratio can be obtained by comparing the out-of-band noise, because the demodulation process before the voice ratio selection has filtered the voice signal, and the out-of-band noise has been filtered out, so the SNR cannot be accurately calculated using this method; The second is the calculation of the average SNR. This method is to calculate the SNR of the entire segment of speech. It needs to buffer the entire segment of speech, calculate the signal energy of the segment and estimate the noise power, and finally calculate the average SNR of the segment of the signal, because This method needs to buffer the whole segment of speech, so the delay is relatively large and does not meet the real-time requirements; the third method is the short-term average signal-to-noise ratio calculation, which is calculated by intercepting part of the speech segment, and this method can meet the real-time requirements. However, there may be errors in the calculation results

Images