基于线性调频Z变换和短波语音通话的飞机类型识别研究

Aircraft type recognition research based on CZT(chirp Z transform) and short-wave speech communication

研究用短波语音通话携带的飞机舱室噪声对飞机类型进行识别的方法。分析了飞机舱室内噪声在短波信道和语音通话干扰下的物理特性,定义了估计语音段的飞机噪声信噪比的公式,提出了自适应的抑制语音增强飞机噪声的模型,通过CZT变换分别提取目标信号不同频段的功率谱密度级特征,并设计了用支持向量机进行分类识别的二叉分类树。对8类现场实测数据进行实验:增强后语音段的平均信噪比提高约22 dB,分类树对语音应答间隔噪声、语音段信号和增强后的信号的平均识别率分别为82.79%,15.25%,50.18%。实验表明:应答间隔噪声可用于飞机类型识别;语音抑制算法带来较大的信噪比和识别率增益,证明语音段蕴含有助于飞机类型识别的重要信息,可为后续的研究奠定基础。

Aircraft information can be transmitted very distant distance by shortwave. The study was executed to recognize aircraft type by using cockpit noise contained in shortwave speech communication. The main interference for cockpit noise is channel and speech noise, which was analyzed. To evaluate cockpit noise in speech segment, SNR （Signal-to-Noise Ratio） equation was defined. To enhance cockpit noise and restrain speech, self-adaptive model was proposed. Power spectrum density level feature for different frequency band was extracted respectively by using CZT. Binary classifier tree based-on support vector machine was designed to recognize aircraft type. Some experiment on eight types＇ aircraft field test data was done to evaluate these methods. The average SNR after enhancing aircraft noise is improved by 22 dB. Classifier output recognition rate for noise between speech response, noise in speech segment and enhanced noise is respectively 82.79%, 15.25%, 50.18%. Experimental results show that, noise between speech response can be used as aircraft type recognition. Speech restrain algorithm brings bigger gain in SNR and recognition rate, which provides existent proofs of aircraft type information in speech segment. Although the recognition rate is still smaller, it can provide reference for later research progress.