基于声纹大数据的防电信诈骗应用研究.docVIP

摘 要 如今，电信网络诈骗发案数量与日俱增，严重危害了人民群众财产安全。随着通信手段的不断发展，通信数据量也越来越大。对于电信诈骗防范系统的实时性和精确性是严重的挑战。传统采用神经网络作为支撑技术的防诈骗系统对于诈骗电话的分辨准确率有限，难以满足对大量声音样本的辨别需求。 本文提出了一种基于声纹大数据的防电信诈骗框架，通过收集并对语音文件进行声纹特征提取形成声纹特征库。在接收到声纹认证请求时，计算需要认证的声纹与声纹特征库中声纹特征的距离从而判断当前认证声纹是否来自于诈骗电话，从而对电信诈骗实施预警。本文所提方法通过并行化提取声纹特征，提高了大数据量情形下程序的处理速度和运行效率。 关键词：电信诈骗；深度学习；声纹识别 Abstract Nowadays, the number of telecommunication network fraud cases is increasing day by day, which seriously endangers the property security of the people. With the continuous development of communication means, the amount of communication data is also increasing. It is a serious challenge to the real-time and accuracy of the telecommunication fraud prevention system. The traditional anti fraud system, which uses neural network as support technology, has limited resolution accuracy for fraudulent phones, and it is difficult to meet the needs of a large number of voice samples. In this paper, we propose an anti fraud framework based on voiceprint big data. The voiceprint feature library is formed by collecting and extracting voiceprint features from voice files. When receiving the voiceprint authentication request, the distance between the voiceprint to be authenticated and the voiceprint feature in the voiceprint feature library is calculated to determine whether the current authenticated voiceprint is from the fraud phone, so as to implement early warning for the telecom fraud. The method proposed in this paper improves the processing speed and running efficiency of the program in the case of large amount of data by parallel extraction of voiceprint features. Key words: Telecommunications Fraud；Deep Learning；Voiceprint Recognition 目 录 TOC \o 1-3 \h \u 第一章 绪论 1 1.1 研究背景 1 1.2 研究现状 1 1.2.1 国外研究现状 1 1.2.2 国内研究现状 2 1.3 研究意义 2 1.4 论文研究内容及结构 3 1.4.1 研究内容 3 1.4.2 论文结构 3 第二章 电信诈骗防范方案 4 2.1 电信诈骗的形式 4 2.2 现有防诈骗技术 4 2.2.1 防诈骗形式 4 2.2.2 防诈骗手段 5 2.3 现有防诈骗技术的缺点 5 2.4 基于声纹的防诈骗技术 6 2.4.1 技术简介 6 2.4.2 应用分析 6 第三章