Page 79 - 201806
P. 79

第 37 卷 第 6 期                                                                        Vol. 37, No.6
             2018 年 11 月                         Journal of Applied Acoustics                 November, 2018


             ⋄ 研究报告 ⋄



                       改进的梅尔倒谱系数在低空飞行器特征

                                               提取中的应用                    ∗





                                              肖寒春 郭俊峰            †   张 丽



                                               (重庆邮电大学通信学院       重庆   400065)

                摘要    梅尔倒谱系数特征提取技术依据人耳的感知特性将声信号从线性频域转换到梅尔域，在语音识别中得
                到广泛应用。该文将梅尔倒谱系数技术用于小型低空飞行器的声信号特征提取中，并针对螺旋桨驱动类的小
                型低空飞行器具有稳定的强谐波特性，对梅尔倒谱系数特征提取中使用的梅尔滤波器进行改进，通过对此类
                谐波处的线性频谱与梅尔谱转换曲线的斜率进行投影替换，提高滤波器对该谐波处信号的感知敏感度。仿真
                结果表明，使用改进的梅尔倒谱系数特征提取方法对小型低空飞行器进行特征提取时，能够得到更低的等误
                识率，并且在低信噪比环境中，改进的梅尔倒谱系数特征提取方法具有更好的抗噪能力。
                关键词     小型低空飞行器，梅尔倒谱系数，特征提取，谐波线
                中图法分类号: TN912.16           文献标识码: A          文章编号: 1000-310X(2018)06-0909-07
                DOI: 10.11684/j.issn.1000-310X.2018.06.012





                 The application of improved Mel-frequency cepstral coefficients technology
                                 in the feature extraction of low-altitude aircraft



                                       XIAO Hanchun     GUO Junfeng     ZHANG Li

                              (Chongqing University of Posts and Telecommunications, Chongqing 400065, China)

                 Abstract  The Mel-frequency cepstrum coefficient (MFCC) feature extraction technology transforms the
                 acoustic signal from the linear frequency domain to the Mel-domain based on the perceptive characteristics
                 of human ear and is widely used in speech recognition. In this paper, MFCC technology is used to extract
                 the acoustic signal features of small low-altitude aircraft, and force on the feature which steady-state strong
                 harmonic of small low-altitude aircraft with propeller driven, to improve the Mel filter for MFCC feature ex-
                 traction. Through the projection and replacement of the slope of the linear spectrum-Mel spectrum conversion
                 curve of this kind of harmonics, the sensitivity of the filter to the signal of the harmonic is improved. Simula-
                 tion results show that using the improved MFCC method can get higher recognition rate, and in the low SNR
                 environment, the improved MFCC feature extraction method has better ability to resist noise.
                 Key words Unmanned aerial vehicle, Mel-frequency cepstrum coefficient, Features extraction, Harmonic
                 lines

             2018-03-09 收稿; 2018-05-01 定稿
             长江学者和创新团队发展计划项目 (IRT1299), 重庆市科委项目 (cstc2013yykfA40010)
             ∗
             作者简介: 郭俊峰 (1990- ), 男, 河南信阳人, 硕士研究生, 研究方向: 模式识别, 数字信号处理。
             † 通讯作者 E-mail: junfeng90@qq.com
   74   75   76   77   78   79   80   81   82   83   84