基于深度学习的管制员语音质量评估  

Assessment of Controller Speech Quality Based on Deep Learning

作　　者：赖鹏 杨倩 Lai Peng;Yang Qian(Air Traffic Control Department of South West Air Traffic,Deyang,China;College of Airport,Civil Aviation Flight University of China,Deyang,China)

机构地区：[1]民航西南空管局空管部,四川德阳 [2]中国民用航空飞行学院机场学院,四川德阳

出　　处：《科学技术创新》2024年第14期1-4,共4页Scientific and Technological Innovation

摘　　要：在航空领域,管制员通过语音通讯获取信息,但是语音信号在经过语音设备的传输后通常会受到干扰或者损伤,该损伤影响着通信的质量。在管制员与飞行员通讯过程中语音质量的优劣直接影响着航空器运行的安全性,为保证通讯过程中语音信号的质量,因此考虑对语音信号进行实时评估。然而主观的评估方法成本高,且主观因素较大,因此考虑通过选取客观语音质量评估方法。基于深度学习的语音质量评估属于无参考的语音质量评估,适应于对通过传输并被记录下的管制语音进行客观评估。With the rapid development of communication technology,the functions and scale of voice communication systems are also undergoing rapid changes.At the same time,various encoding and decoding technologies are emerging one after another.Different types of encoding and decoding technologies and transmission methods can cause varying degrees of damage to speech quality.The quality of voice during communication between controllers and pilots directly affects the efficiency of communication between both parties and the safety of aircraft operation.To ensure the quality of voice signals during communication,this article considers real-time evaluation of voice signals.However,subjective evaluation methods have high costs and significant subjective factors.Therefore,it is considered to select objective speech quality evaluation methods for evaluating controlled speech.Deep learning based speech quality assessment is a reference free speech quality assessment method that is precisely suitable for objective evaluation of directly recorded controlled speech.

关 键 词：语音质量评价 深度学习 神经网络 空管语音 

分 类 号：TN912.3[电子电信—通信与信息系统] TP181[电子电信—信息与通信工程]