Page 107 - 《应用声学》2020年第1期

第 39 卷 第 1 期 陈建新等： 基于子带信号瞬时频率的特征提取及其在车型分类中的应用 103

95
based on Shamma’s model[J]. Journal of Applied Acous-
90 tics, 2011, 30(6): 407–417.
[3] Nakagawa S, Wang L, Ohtsuka S. Speaker identification
85
and verification by combining MFCC and phase informa-
юᆸဋ/% 80 tion[J]. IEEE Transactions on Audio, Speech and Lan-
guage Processing, 2012, 20(4): 1085–1095.
75
phase spectrum in human listening tests[J]. Speech Com-
70 [4] Paliwal K K, Alsteris L D. On the usefulness of STFT
IF(Mel) munication, 2005, 45(2): 153–170.
65 MFCC
[5] Duarte M F, Hu Y H. Vehicle classification in distributed
60 sensor networks[J]. Journal of Parallel and Distributed
-5 0 5 10 15 20 Computing, 2004, 64(7): 826–838.
η٪උ/dB
[6] 王敏, 赵鹤鸣. 基于多带解调分析和瞬时频率估计的耳语音话
图 4 不同信噪比下两种特征分类性能比较 (数据集 1) 者识别 [J]. 声学学报, 2010, 35(4): 471–476.
Fig. 4 Comparison of two features classification perfor- Wang Min, Zhao Heming. Whispered speaker identifi-
mance under different signal to noise ratio(Data Set 1) cation based on multiband demodulation analysis and in-
stantaneous frequency estimation[J]. Acta Acustica, 2010,
90 35(4): 471–476.
[7] Li Q, Huang Y. An auditory-based feature extraction algo-
80
rithm for robust speaker identification under mismatched
70 conditions[J]. IEEE Transactions on Audio, Speech and
Language Processing, 2011, 19(6): 1791–1801.
юᆸဋ/% 60 [8] Shamma S, Klein D. The case of the missing pitch tem-
plates: how harmonic templates emerge in the early audi-
50
tory system[J]. Journal of the Acoustical Society of Amer-
40
IF(Mel) ica, 2000, 107(5): 2631–2644.
MFCC
30 [9] Patel T B, Patil H A. Cochlear filter and instantaneous
frequency based features for spoofed speech detection[J].
20
-5 0 5 10 15 20 IEEE Journal of Selected Topics in Signal Processing,
η٪උ/dB 2017, 11(4): 618–631.
[10] 顾明亮, 夏玉果, 杨亦鸣. 支持矢量机的汉语声调识别 [J]. 声
图 5 不同信噪比下两种特征分类性能比较 (数据集 2)
学技术, 2007, 26(6): 1186–1190.
Fig. 5 Comparison of two features classification perfor-
Gu Mingliang, Xia Yuguo, Yang Yiming. Support vec-
mance under different signal to noise ratio(Data Set 2) tor machine based Chinese tone recognition[J]. Technical
Acoustics, 2007, 26(6): 1186–1190.
6 结论 [11] 刘旭, 夏金东, 弓乐, 等. 射频信号在超声检测缺陷识别中的
应用研究 [J]. 机械工程学报, 2002, 38(4): 84–87.
在基于声信号的车型分类研究中，本文提出了 Liu Xu, Xia Jindong, Gong Yue, et al. Study on the ap-
一种基于耳蜗滤波器组的 IF特征提取算法，并将该 plication of radio-frequency signal in flaw classification of
ultrasonic testing[J]. Journal of Mechanical Engineering,
特征与对数能量相结合，通过可分性测度和分类实 2002, 38(4): 84–87.
验表明所提取特征可以有效地实现车型分类并且 [12] 张浩然, 韩正之, 李昌刚. 基于支持向量机的非线性模型预测
抗噪性良好。本文实验也验证了相位信息同样能够 控制 [J]. 系统工程与电子技术, 2003, 25(3): 330–334.
Zhang Haoran, Han Zhengzhi, Li Changgang. Support
反映声目标的重要信息，在声目标识别上有巨大应 vector based nonlinear model predictive control[J]. Sys-
用潜力。 tems Engineering Electronics, 2003, 25(3): 330–334.
[13] 毕超, 冯玉田, 李园辉, 等. 基于 HHT 的声频传感器车辆分类
参 考 文 献 识别 [J]. 电声技术, 2016, 40(6): 48–52.
Bi Chao, Feng Yutian, Li Yuanhui, et al. Audio sensor
[1] Aljaafreh A, Dong L. An evaluation of feature extraction
vehicle classification recognition based on HHT[J]. Audio
methods for vehicle classification based on acoustic sig-
Engineering, 2016, 40(6): 48–52.
nals[C]. Networking, Sensing and Control (ICNSC), 2010
[14] Wang K, Wang R, Feng Y, et al. Vehicle recognition
International Conference on. IEEE, 2010: 570–575.
in acoustic sensor networks via sparse representation[C].
[2] 陈克安, 伍莹, 杨立学. 基于 Shamma 模型的车辆噪声听觉谱
IEEE International Conference on Multimedia and Expo
特性与目标分类 [J]. 应用声学, 2011, 30(6): 407–417.
Workshops. IEEE, 2014.
Chen Ke’an, Wu Ying, Yang Lixue. Auditory spectral
[15] 宋知用. MATLAB 在语音信号分析与合成中的应用 [M]. 北
characteristics and target classification for vehicle noise
京: 北京航空航天大学出版社, 2013.