作者：韩威1,2, 周松斌1, 刘忆森1, 李昌1, 刘伟鑫1,2
作者单位：1. 广东省智能制造研究所广东省现代控制技术重点实验室, 广东 广州 510070;
2. 广东工业大学, 广东 广州 510006
关键词：麦克风阵列;稀疏半非负矩阵分解;声学检测;罐装食品真空度检测
摘要：
声音频谱峰值法被广泛应用于罐装食品真空度检测领域，但是当检测环境出现声音强度较大且与罐盖振动产生的声音的频段相同的噪声时，该方法可能做出误判。为此，提出声学阵列法：由麦克风阵列采集多路混合声信号，采用稀疏半非负矩阵分解从混合声信号中分离出干净的罐盖振动产生的声音，再利用声音频谱峰值法判断真空度是否合格。该文研究稀疏半非负矩阵分解的数学模型，并且推导求解稀疏半非负矩阵分解的迭代优化函数。实验结果表明，无噪声环境下，声音频谱峰值法和声学阵列法的真空度检测结果均准确，但在噪声环境下，声音频谱峰值法出现误判时，声学阵列法仍能做出准确判断。
Online detection for vacuum of canned food based on microphone array
HAN Wei1,2, ZHOU Songbin1, LIU Yisen1, LI Chang1, LIU Weixin1,2
1. Guangdong Key Laboratory of Modern Control Technology, Guangdong Institute of Intelligent Manufacturing, Guangzhou 510070, China;
2. Guangdong University of Technology, Guangzhou 510006, China
Abstract: The method based on the acoustic spectral peak is widely used in the field of inspecting the vacuum of canned food. However, this method may make a misjudgment due to the high intensity noise with the same frequency band of sound produced by the vibration of the lid of canned food. For this reason, this paper proposes an acoustic array method: a microphone array is used to collect multi-channel mixed sound signal, and then the clean sound produced by the vibration of tank cover is separated from mixed sound signal by sparse semi-nonnegative matrix factorization, determining whether the vacuum is qualified depends on the peak of the sound spectrum in the subsequent handling. This paper studied the mathematical model of semi-nonnegative matrix factorization (SSNMF), and derived the functions for solving the SSNMF by iterative strategy. Experimental results show that the detection results from the method based on both the acoustic spectrum peak and acoustic array method are correct under the noiseless condition. However, in the noise environment, the acoustic array method could still make a correct judgment when the method based on the acoustic spectrum peak provides an erroneous estimation.
Keywords: microphone array;sparse semi-nonnegative matrix factorization;acoustic detection;detection for vacuum of canned food
2019, 45(7):128-133 收稿日期: 2018-09-10;收到修改稿日期: 2018-11-08
基金项目: 国家自然科学基金资助项目（61803107）；广东省科技计划资助项目（2016B090918061）；广州市科技计划资助项目（201803020025）
作者简介: 韩威(1987-),男,湖北荆门市人,助理研究员,博士,主要从事传感技术与在线无损检测技术研究
参考文献
[1] 张振祥, 张立俊, 陈连才, 等. 罐头真空度检测仪器的研究[J]. 现代商检科技, 1996, 6(4):5-7
[2] 刘南平, 崔雁松, 李佳, 等. 双控头电涡流法食品罐头真空度测量仪[J]. 传感技术学报, 2005, 18(1):129-131
[3] 刘南平, 李文英, 崔雁松, 等. 食品罐头真空度无损检测仪的研制[J]. 传感器与微系统, 2004, 23(12):40-42
[4] 干蜀毅, 朱武, 陈长琦, 等. 电涡流法检测罐头真空度的探头研制和仪器智能化[J]. 真空科学与技术, 2002, 22(5):326-328
[5] GARCÍAMARTÍN J, GÓMEZGIL J, VÁZQUEZSÁNCHEZ E. Non-destructive techniques based on eddy current testing[J]. Sensors, 2011, 11(3):2525-2565
[6] 付江云, 刘怡俊, 陈靖宇, 等. 基于声学原理的啤酒瓶在线内压力检测系统设计[J]. 计算机测量与控制, 2013, 21(9):2362-2368
[7] 张业伟, 徐全金, 侯鹏飞. 基于DSP的啤酒瓶封盖密封性在线检测系统的设计[J]. 智慧工厂, 2011(7):121-122
[8] 陈靖宇, 付江云, 谢振南, 等. 基于实时滤波的瓶盖密封性测试系统设计[J]. 计算机测量与控制, 2012, 20(4):951-954
[9] SUGIMOTO K, SUGIMOTO T, UTAGAWA N, et al. The defect detection algorithm that combined spectrum entropy with vibrational energy ratio for acoustic inspection method[J]. Acoustical Society of America Journal, 2017, 141(5):3831
[10] 金龙, 朱振池, 刘千令, 等. 基于激振声学的物体内部缺陷检测装置[J]. 电子技术应用, 2015, 41(2):45-47
[11] ELFORJANI M, SHANBR S. Prognosis of bearing acoustic emission signals using supervised machine learning[J]. IEEE Transactions on Industrial Electronics, 2018, 65(7):5864-5871
[12] 张海澜. 理论声学[M]. 北京:高等教育出版社, 2012:168-171.
[13] 黄光周, 蔡德全. 食品罐头内真空度的无损检测[J]. 华南工学院学报(自然科学版), 1987, 15(4):13-17
[14] CHRIS D, TAO L, MICHAEL I J. Convex and semi-nonnegative matrix factorizations[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2010, 32(1):45-55
[15] ALBERTO P,CARAZO J M, KIEKO K, et al. Nonsmooth Nonnegative Matrix Factorization (nsNMF)[J]. IEEE Transa-ctions on Pattern Analysis and Machine Intelligence, 2006, 28(3):403-414
[16] 周治宇, 陈豪. 盲信号分离技术研究与算法综述[J]. 计算机科学, 2009, 36(10):16-20
