臺灣博碩士論文加值系統

當新生兒出生時，還未學習如何說話時，都是利用哭聲與臉部表情來傳達訊息。對新生兒而言，哭聲可以作為嬰兒與外界最主要的溝通方式。當新生兒哭鬧時，新手爸媽常常不能了解新生兒的需求，因此，如果可以準確判斷出新生兒不同的哭聲所代表的含意，那麼就能夠了解嬰兒情緒以及生理需求的各項變化與需求。
本論文提出一套自動化辨識嬰兒哭聲的系統，本研究轉換哭聲訊號由時間域轉成頻率域，使用2維聲譜圖作為輸入的樣本，透過深度學習中卷積深度神經網路架構(Convolutional Neuron Network, CNN)計算後，根據不同日齡的新生兒成功分辨出三種模式，包含飢餓、疼痛以及想睡覺，比起一般的神經網路能有更短的執行時間。

Crying is a way which infants express their needs to their parents. In general, parents often feel worried and anxious when infant crying.
For realizing the reason of baby crying, this paper use convolutional neuron networks to train our own cry data and development an automated identification cries model. This system assists parents to understand the demand of the infants. Transfer audio signal from time domain to frequency domain and use two-dimensional spectrogram as input. After Convolutional neuron networks computing, this paper have successfully distinguished three categories based on different baby birthday. Convolutional neuron networks does not extract feature and additional pre-processing. Compared with general machine learning network it with much less query time.

摘要 i
ABSTRACT ii
目錄 iii
表目錄 v
圖目錄 vi
第一章 緒論 1
1.1 研究動機與目的 1
1.2研究目的 2
1.3 文獻探討 2
1.4 研究方法 3
1.5 章節大鋼 4
第二章 相關理論 6
2.1 聲音的產生 6
2.1.1 聲音振幅(Magnitude) 7
2.1.2音調(Pitch) 8
2.1.3 音色(Timbre) 9
2.2 短時距傅立葉轉換(SHORT-TIME FAST FOURIER TRANSFORM, STFT) 9
2.3 共振峰(FORMANT) 14
2.4 聲譜圖(SPECTROGRAM) 16
2.5 深度學習(DEEP LEARNING) 17
2.6卷積神經網路(CONVOLUTIONAL NEURON NETWORKS) 19
第三章 研究方法 28
3.1 系統架構 28
3.2 訊號前處理 30
3.2.1 音框化(framing) 31
3.2.2 端點偵測(End-point Detection，EPD) 32
3.3 短時距傅立葉轉換 34
3.4 訓練階段 35
3.5 測試階段 35
第四章 實驗結果與討論 36
4.1 實驗環境與設計 36
4.2 實驗資料庫 36
4.3利用K-fold交叉驗證法之新生兒哭聲模式分類結果 37
4.4 與SVM的方法做比較 39
4.5 本論文方法與未區分月齡方法做比較 39
第五章 結論 41
參考文獻 42

[1]S. F Carbaugh, "Understanding shaken baby syndrome, " Adv Neonatal Care, 2004, Available：http://www.medscape.com/viewarticle/478153_3.
[2]Crying in infancy, Available: http://pennstatehershey.adam.com/content.aspx?productId=112&pid=1&gid=002397.
[3]M. Silva, B. Mijovic, Bea R.H. Van den Bergh, K. Allegaert, J.M. Aerts, S. Van Huffel, and D. Berckmans, "Decoupling between fundamental frequency and energy envelope of neonate cries," Early Human Development, vol. 86, pp. 35-40, 2010.
[4]萊恩。《嬰幼兒發展》。五南圖書出版。
[5]王小川(2008)。《語音訊號處理》。臺北：全華圖書。
[6]D. Huron, "The ramp archetype and the maintenance of passive auditory attention," Music Perception, vol. 10, pp. 83-91, 1992.
[7]J. S. R. Jang. Audio Signal Processing and Recognition, Available: http://mirlab.org/jang/books/audioSignalProcessing/.
[8]S. Z. Li, "Content-based audio classification and retrieval using the nearest feature line method," IEEE Transactions on Speech and Audio Processing, vol. 8, no. 5, pp. 619-625, 2000.
[9]M. Liu and C. Wan, "A study on content-based classification and retrieval of audio database," International Symposium on Database Engineering and Applications, pp. 339-345, 2001.
[10]R. J. Mammone, Z. Xiaoyu, and R. P. Ramachandran, "Robust speaker recognition: a feature-based approach," IEEE Signal Processing Magazine, vol. 13, no. 5, pp. 58-71, 1996.
[11]J. W. Cooley, and J. W. Tukey, "An algorithm for the machine calculation of complex Fourier series," Mathematics of Computation, vol. 19, pp. 297-301, 1965.
[12]王士元、彭剛(2007)。《語言、與音與技術》。香港城市大學出版社。
[13]陳用佛、鄒濬智、沈文聖(2013)。《破案關鍵: 指紋、毛髮、血液、DNA，犯罪現場中不可不知的鑑識科學》。獨立作家出版。
[14]KRIZHEVSKY, Alex; SUTSKEVER, Ilya; HINTON, Geoffrey E. " Imagenet classification with deep convolutional neural networks, " Advances in neural information processing systems, pp. 1097-1105, 2012
[15]張斐章、張麗秋(2007)。《類神經網路》。臺灣東華書局股份有限公司。
[16]Crying in infancy, Available:
http://www.bnext.com.tw/article/view/id/38923。
[17]Y. LeCun, L. Bottou, Y. Bengio, and P. Haffner. "Gradient-based learning applied to document recognition, " Proceedings of the IEEE, vol. 86, pp.2278–2324, 1998.
[18]http://www.wikiwand.com/zh-hk/%E5%8D%B7%E7%A7%AF%E7%A5%9E%E7%BB%8F%E7%BD%91%E7%BB%9C。
[19]Nitish Srivastava, Geoffrey Hinton, Alex Krizhevsky, Sneddon, Ilya Sutskever, Ruslan Salakhutdinov . " Dropout: A Simple Way to Prevent Neural Networks from Overfitting," Journal of Machine Learning Research archive, vol.15 Issue 1, pp.1929-1958, 2014.
[20]Hinton, G., Srivastava, N., Krizhevsky, A., Sutskever, I.,and Salakhutdinov, R. " Improving neural networks by preventing co-adaptation of feature detectors, " Computing Research Repository, pp.1207.0580, 2012.
[21]Crying in infancy, Available:
https://penolove.gitbooks.io/deep-learning-tutourial/content/Supervise%20learning/softmax%20regression.html。
[22]M. A. Ruíz Díaz, C. A. Reyes García, L. C. Altamirano Robles, J. E. Xalteno Altamirano, and A. Verduzco Mendoza, "Automatic infant cry analysis for the identification of qualitative features to help opportune diagnosis, "Biomedical Signal Processing and Control, vol. 7, pp. 43-49, 2012.
[23]Campbell, William M., Douglas E. Sturim, and Douglas A. Reynolds. "Support vector machines using GMM supervectors for speaker verification." Signal Processing Letters, IEEE, pp.308-311, 2006.
[24]Abdulla, Waleed H., David Chow, and Gary Sin. "Cross-words reference template for DTW-based speech recognition systems," Conference on Convergent Technologies for the Asia-Pacific Region, vol. 4. IEEE, 2003.
[25]Chang, C. Y., Hsiao, Y. C., & Chen, S. T. "Application of Incremental SVM Learning for Infant Cries Recognition," Network-Based Information Systems, pp. 607-610, 2015.
[26]P. Dhanalakshmi, S. Palanivel, and V. Ramalingam, "Classification of audio signals using SVM and RBFNN, " Expert Systems with Applications, vol. 36, no. 3, pp. 6069-6075, 2009.
[27]Y. H. Yang, Y. C. Lin, Y. F. Su, and H. H. Chen, "A regression approach to music emotion recognition," IEEE Transactions on Audio, Speech, and Language Processing, vol. 16, no. 2, pp. 448-457, 2008.
[28]D. C. Park, "Classification of audio signals using fuzzy c-means with divergence-based kernel," Pattern Recognition Letters, vol. 30, no. 9, pp. 794-798, 2009.
[29]T. Li and M. Ogihara, "Toward intelligent music information retrieval," IEEE Transactions on Multimedia, vol. 8, no. 3, pp. 564-574, 2006.
[30]G. Tzanetakis and P. Cook, "Musical genre classification of audio signals," IEEE Transactions on Speech and Audio Processing, vol. 10, no. 5, pp. 293-302, 2002.
[31]H. Lee, P. Pham, Y. Largman and A. Ng. " Unsupervised Feature Learning for Audio Classification Using Convolutional Deep Belief Networks," In Proc. Neural Information and Processing System, 2009
[32]Graves, Alex, and Navdeep Jaitly. "Towards end-to-end speech recognition with recurrent neural networks," Proceedings of the 31st International Conference on Machine Learning, vol. 14, pp. 1764-1772, 2014.
[33]DENG, Li, et al. " Binary coding of speech spectrograms using a deep auto-encoder, " In: Interspeech, pp. 1692-1695, 2010..
[34]Chang, C. Y., Chang, C. W., Kathiravan, S., Lin, C., & Chen, S. T. "DAG-SVM based infant cry classification system using sequential forward floating feature selection," Multidimensional Systems and Signal Processing, 1-16,2016
[35]R. Kohavi, "A study of cross-validation and bootstrap for accuracy estimation and model selection, "Proceedings of the 14th international joint conference on Artificial intelligence, pp. 1137-1143, 1995.