Please use this identifier to cite or link to this item: http://hdl.handle.net/11455/4872
標題: 智慧型嬰兒監護系統之自動情緒辨識與咳嗽偵測
Automatic emotion classification and cough detection in intelligent baby monitoring systems
作者: 許家豪
Shiu, Jia-Hau
關鍵字: dynamic time warping
動態時軸校正
sliding window
neural network
mel frequency cepstrum
filter of bank
bandpass filter
baby cry
automatic cough detection.
移動視窗
類神經網路
梅爾倒頻譜
濾波器組
帶通濾波器
嬰兒哭聲
自動咳嗽偵測
出版社: 通訊工程研究所
引用: 中文部份 [1] 許積德,“嬰兒哭鬧探緣由”,維普資訊,http://engine.cqvip.com/。 [2] 劉穎,于志丹,“淺談嬰兒哭鬧200例的病因分析”,齊齊哈爾醫學院學報第二十七卷第三期,2006。 [3] 高天霽,“嬰幼兒慢性咳嗽92例病因分析”,河北醫藥第三十五卷第十一期,2007。 西文部份 [4] M.J. Doherty*, L.J. Wang*, S. Donague*, M.G. Pearson*, P. Downs**, S.A.T. Stoneman**, J.E. Earis*, “The acoustic properties of apsaicin-induced cough in healthy subjects”, ERS Journals ISSN 0903-1936, 1997. [5] Orion F. Reyes-Galaviz and Carlos Alberto Reyes-Garcia, “A System for Processing of Infant Cry to Recognize Pathologies in Recently Born Babies with Neural Network” ISCA Archive,2004. [6] Sandra E.Barajas-Montiel,Carlos A. Reyes-Garcia, Emilio Arch-Tirado and Arch-Tirado,“Improving Baby Caring with Automatic Infant Cry Recognition”, ICCHP 2006, LNCS pp.691-698,2006. [7] Sergio Matos, Surinder S. Birring, Ian D. Pavord, and David H. Evans, “Detection of Cough Signals in Continuous Audio Recordings Using Hidden Markov Models”, IEEE TRANSACTIONS ON BIOMEDICAL ENGINEERING, VOL.53, 2006. [8] M. Guarino, P.Jans, A. Costa, J-M. Aerts and D. Berckmans, “Field test of algorithm for automatic cough detection in pig houses”, Computers and Electronics in Agriculture, VOL.62, pp.22-88, 2008. [9] J. Korpa´sˇ*, J. Sadlonˇova´†, M. Vrabec, “Methods of Assessing Cough and Antitussives in Man”Pulmonary Pharmacology, Vol 9, pp.373-377, 1996. [10] Hiew, Y., Smith, J., Earis, J., Cheethma, B., Woodcock, A, “Dsp algorithm for cough identification and counting”, In: Proc. ICASSP ’02, Orlando, Florida, pp.3888–3891, 2002. [11] Van Hirtum, A., Berckmans, D, “Automated recognition of spontaneous versus voluntary cough”, Med. Eng. Phys. 24, pp.541–545, 2002. [12] Van Hirtum, A., Berckmans, D, “Fuzzy approach for improved recognition of citric acid induced piglet coughing from continuous registration”, J. Sound Vibrat. 266, pp.667–686, 2003a. [13] Van Hirtum, A., Berckmans,D, “Intelligent free field cough sound recognition” In: Proc. ICONS ’03, Faro, Portugal, pp. 453–458, 2003b. [14] Wathes, C.M., Jones, J.B., Kristensen, H.H., Jones, E.K.M., Webster, A.J.F., “Aversion of pigs and domestic fowl to atmospheric ammonia. Trans”,ASAE 45(5), pp.1605–1610,2002. [15] Sung-Hwan Shin, Takeo Hashimoto, and Shigeko HatanoAutomatic “Detection System for Cough Sounds as a Symptom of Abnormal Health Condition”, IEEE 2008. [16] V. Exadaktylos, M. Silva, J.-M. Aerts, C.J. Taylor and D. Berckmans,“Real-time recognition of sick pig cough sounds”,Computers and Electronics in Agriculture,VOL.63,pp.207-214, 2008. [17] Sara Ferrari, Mitchell Silva, Marcella Guarino, Jean Marie Aerts and Daniel Berckmans, “Cough sound analysis to identify respiratory infection in pigs”, Computers and Electronics in Agriculture, VOL.64, pp.318-325, 2008. [18] Orion F. Reyes-Galaviz, Emilio Arch Tirado, and Carlos Alberto Reyes-Garcia,“Classification of Infant Crying to Identify Pathologies in Recently Born Babies with ANFIS”,Lecture Notes in Computer Science,Vol.3118,2004 [19] Sandra E. Barajas-Montiel, Carlos A. Reyes-Garcia, Emilio Arch-Tirado and Mario Mandujano,“Improving Baby Caring with Automatic Infant Cry Recognition”,Lecture Notes in Computer Science, Vol.4061,2006. [20] Cong Phuong Nguyen, Thi Ngoc Yen Pham and Castelli Eric, “Toward a sound Analysis System for Telemedicine”, Lecture Notes in Computer Science, Vol.3614, 2005. [21] Yuki Mima and Kaoru Arakawa, “Cause Estimation of Younger Babies'' Cries from the Frequency Analyses of the Voice”, ISPACS 2006.
摘要: 在智慧型嬰幼兒監護系統中,需要利用即時擷取的音訊信號,自動偵測被監護者的身心狀態,以提昇監護系統的效能。就聲音訊號而言,如果能夠依據所取得的哭聲訊號狀態,能有助於提供適當的照護行動,此外,考量夜咳是嬰兒常見的病徵,如果能自動偵測並記錄夜咳次數,將有助於醫生做為疾病診治與用藥之前的參考,並降低父母親的照顧負擔。 在情緒辨識的部份,我們將從哭聲訊號,判別兩種情緒狀態:痛與餓。利用MFCC參數作為哭聲訊號的特徵參數,以及KNN與倒傳遞類神經網路做為分類方法,我們所建購的情緒偵測系統可以達到89%的準確率。 在咳嗽自動偵測的部份,我們將提出根據動態時軸校正(Dynamic Time Warping,DTW)處理兩筆長短不一的咳嗽訊號特徵向量之自動偵測方法。有別於相關研究[7]中,需要額外的端點偵測處理,將輸入訊號中含有咳嗽聲音的部份切割以後,再進行DTW偵測處理,在我們所提出的方法中,利用移動視窗(sliding window)取得不同段落的訊號,進行DTW偵測處理,即可免除額外的端點偵測。為了降低計算複雜度,我們將利用相鄰移動視窗內資料的重複性,針對相似度高的兩個相鄰視窗,以複製前一視窗的DTW比對路徑,簡化下一視窗的DTW路徑搜尋,針對相似度低的部份,則利用前一視窗的DTW比對路徑,計算出下一視窗的DTW比對距離的上限值。實驗結果顯示,利用上述之簡化處理,可以降低65%計算複雜度,同時達成96.7%正確偵測率。
In intelligent baby monitor systems, there needs methods to get instaneous ststus of signal, which can promote monitor system's efficiency by automatic detect watchdog's status. If we can detect infant's emotion and cough numbers automaticly in signal process section, which can help doctors to be a reference befor taking a medicine and disease diagnose, and it also helps parents to reduce caring burden. In emtional classification, there majors judge two kinds of cries of infant, pain and hunger. We get crying signal feature by using MFCC parameters, and we use features in KNN and ANN classification systems. In our emotional system, the infant's crying correction of classification reachs 89%. In automatic cough detection, we address automatic detection method to solve two cough signals of feacture vector with differenct length by dynamic time warping (DTW). Our method is different to [7], which needs additional end-point detection to cut off cough signal from input signal and then using cough signal to carry out DTW algorithm. In our method, we using sliding window to catch different signal to carry out DTW algorithm, which avoids additional end-point detection. DTW method has computation complexity problem. In order to reduce computation complexity of DTW, we use data redundances between adjacent sliding windows. If the difference is low between two adjacent sliding windows, the next sliding window can continue using optimum path of front sliding window to compute the DTW distance. On the other hand, the difference is high between two adjacent sliding windows, the next window can use path of front sliding window to compute the opper-bound of DTW distance. Simulation results show that the correct detection rate of our automatic cough detection method is 96.7% and reduce 65% operation complexity.
URI: http://hdl.handle.net/11455/4872
其他識別: U0005-2907200917572400
文章連結: http://www.airitilibrary.com/Publication/alDetailedMesh1?DocID=U0005-2907200917572400
Appears in Collections:通訊工程研究所

文件中的檔案:

取得全文請前往華藝線上圖書館



Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.