摘要
本研究结合当前人工智能的最新技术进展,采用基于识别广播语音关键词的方法,设计了基于广播关键词进行自动无线电频谱信号管控的方法,具有成本低、效率高、可连续工作等特点。另外,针对无线电信号噪音大、非法广播出现时间不确定、且非法广播种类范围较广的特点,本研究提供了针对不同类型的违规广播进行识别的定制功能;同时,不同于主要网络公司提供的语音识别模型需要联网使用的特点,本研究可以在各类平台上内置模型,离线进行使用。
This research combines the latest technological advances of artificial intelligence, and adopts a method based on identifying broadcast speech keywords to automatically carry out radio spectrum control It has the characteristics of low cost, high efficiency and can be used for continuous operation In addition, in view of the high noise of radio signals, the uncertain time of illegal broadcasts, and the wide range of illegal broadcasts, this study provides customized functions for identifying different types of illegal broadcasts Furthermore, it is applicable without internet connection, especially suitable for systems emphasising security and privacy This study can be adapted for various platforms and used offline.
引文
[1]杨富雄.如何在日常监测中有效识别“黑广播”[J].中国无线电,2016(6):18-20.
[2]桑宇,沈建强,李浩,邱荣.基于信号特征的“黑广播”监测方法探究[J].中国无线电,2015(12),39-40.
[3]赵长青,邸娜,王旖旎.基于DM642的水印检测在音频广播系统中的应用[J].电声技术,2015(4):40-43.
[4]黄寅,赵长青,程伟,王旖旎,吴小茜,陆群.音频水印嵌入系统的设计与实现[J].广播与电视技术,2015(6):120-122.
[5]Saon,George,and Michael Picheny.Recent advances in conversational speech recognition using convolutional and recurrent neural networks.IBM Journal of Research and Development.2017(61):1-4.