影视动画配音节奏特征自动提取系统设计  

Design of automatic extraction system for dubbing rhythm features of film and television animation

作　　者：章瑾[1] 冯平 ZHANG Jin;FENG Ping(College of Post and Telecommunications of WIT,Wuhan 430070,China)

机构地区：[1]武汉工程大学邮电与信息工程学院,湖北武汉430070

出　　处：《现代电子技术》2020年第18期59-63,共5页Modern Electronics Technique

基　　金：国家自然科学基金项目(20873097)。

摘　　要：配音节奏特征对于影视动画人物理解与情绪表达尤为重要,基于此,设计影视动画配音节奏特征的自动提取系统。配音采集模块采用TLV320AIC23音频接口芯片,该芯片内部集成A/D和D/A转换器,可对采集的影视动画配音音频进行模数转换,将获取的模拟信号传输至节奏特征提取模块。节奏特征提取模块以TMS320VC5509 DSP芯片为核心,采用优化后的哈佛结构,利用配音节奏特征提取算法提取影视动画配音节奏特征。配音节奏特征提取算法中,通过快速傅里叶变换确定复数域信息,确定发音点检测信号,以及各帧发音点检测信号函数的自相关函数。基于上下文的节拍周期推导算法推导配音片段节拍周期。基于发音点检测信号函数,通过HMM模型可推算节拍出现的详细时间点并进行节拍跟踪,确定节拍速度特征。实验结果显示,所设计系统可准确划分影视动画配音情绪类型,且能耗较低。The dubbing rhythm features are particularly important for the understanding and emotional expression of the characters in the film and television animation.Based on this,an automatic extraction system for the dubbing rhythm features of film and television animation is designed.TLV320AIC23 audio interface chip is used in the dubbing acquisition module,and A/D converter and D/A converter is internally integrated in the chip,which can execute the analog⁃to⁃digital conversion of the collected dubbing audio of the film and television animation,and transmit the obtained analog signal to the rhythm feature extraction module.The rhythm feature extraction module takes TMS320VC5509 DSP chip as its core,adopts the optimized Harvard structure,and extracts the dubbing rhythm features of the film and television animation by means of the dubbing rhythm feature extraction algorithm.In the algorithm of dubbing rhythm feature extraction,the complex number field information is determined by the fast Fourier transform,so as to confirm the detection signal of the pronunciation point and make sure the autocorrelation function of the pronunciation point detection signal function of each frame.The beat cycle of the dubbing segment is deduced according to the contextual beat cycle derivation algorithm.According to the function of pronunciation point detection signal,the detailed time points where the beat appears can be deduced with HMM model to track the beats and determine the beat speed features.The experimental results show that the designed system can accurately classify the dubbing emotion types of film and television animation,and has low energy consumption.

关 键 词：影视动画 配音采集 节奏特征提取 发音点检测 系统设计 节拍跟踪 

分 类 号：TN911.23-34[电子电信—通信与信息系统] TP391[电子电信—信息与通信工程]