臺灣博碩士論文加值系統

人工智慧是從二十世紀中期在電腦及電腦程式語言創造出來後才開始的研究。電腦計算能力與平行處理使得人工智慧在某些特殊的領域優於人類。許多的類神經網路讓電腦能從資料學到規則。電腦每個迴圈都會學習一些，經過幾千幾百萬的迴圈他們就會學習到規則並做的比人好。
我們使用人工智慧的技術在下列應用：骨導麥克風及老年智慧手環。
骨導麥克風：
骨導麥克風是基於說話者頭骨的振動來產生語音，在使用骨導麥克風做傳輸語音時它們比正常的氣導麥克風表現出更好的抗噪能力。但因為骨導麥克風只能擷取語音信號的低頻部分，所以它們的頻率響應與氣導麥克風的頻率響應差異很大。當用骨導麥克風代替氣導麥克風時，我們可以在噪音抑制方面取得令人滿意的結果，但由於固體振動的性質，語音品質和理解度可能會降低。 骨導麥克風和氣導麥克風的不一致特性也會影響語音辦識的性能，並且使用骨導麥克風的語音數據重新創建新的語音辦識系統是不可行的。在這項研究中，我們提出了一種新穎的深度去噪自動編碼器方法來橋接骨導麥克風和氣導麥克風，以提高語音品質和可理解度，可以直接使用目前的語音辦識系統而不需重新創建新系統。實驗結果顯示，深度去噪自動編碼器的方法可以有效提高語音品質和可理解度在一些標準化評估指標。此外，我們提出的系統可以顯著提高語音辦識性能，在無噪音的條件下，相對字元錯誤率降低了48.28％（從14.50％降至7.50％）。在實際噪音環境下（噪音聲壓從61.7 dBA到73.9 dBA），我們提出的使用骨導麥克風加深度去噪自動編碼器優於氣導麥克風，相對字元錯誤率（骨導麥克風：9.13％和氣導麥克風：58.75％）減少84.46％。
老年智慧手環：
現代醫學使更多的人口能夠生存超過65歲。長者照護已成為唯一最相關的全球挑戰之一。有許多相關議題與監測及管理老年人的日常活動。在這個研究中，我們將檢測使用先進活動傳感平台在監控長者日常活動中的活動水平。它將有助於了解個別老年人的生活方式促進其安全並提高生活素質經由量測身體上的力量及獨立性。我們調查了可穿戴設備的要求用於老年人的技術，採用慣性測量傳感器。我們建議的系統包括每個老年人的可穿戴設備，物聯網接收器環境，智能警報，機器學習算法，以及與使用遠程網路的應用程序處理介面。我們提出的慣性測量傳感器應用的主要包括精確測量個人身體活動水平。

Artificial Intelligence (AI) study starts from the middle of 20 century after computers and computer languages are created. Computer calculation power and parallel processing make AI superior to human in specific domain. Many different neural networks make computers learn the rules from data. Computers learn a little in every loop. After thousands and millions of loops, they finally learn the rules and perform better than human.
We use AI technologies in the following applications: bone-conducted microphone and senior smart band.
Bone-Conducted Microphone (BCM):
BCMs capture speech signals based on the vibrations of the talker’s skull, and they exhibit better noise-resistance capability than normal Air-Conducted Microphones (ACMs) when transmitting speech signals. Because BCMs only capture the low-frequency portion of speech signals, their frequency response is quite different from that of ACMs. When replacing an ACM with a BCM, we may obtain satisfactory results with respect to noise suppression, but the speech quality and intelligibility may be degraded owing to the nature of the solid vibration. The mismatched characteristics of BCM and ACM can also impact the Automatic Speech Recognition (ASR) performance, and it is infeasible to recreate a new ASR system using the voice data from BCMs. In this study, we propose a novel Deep-Denoising AutoEncoder (DDAE) approach to bridge BCM and ACM in order to improve speech quality and intelligibility, and the current ASR could be employed directly without recreating a new system. Experimental results first demonstrate that the DDAE approach can effectively improve speech quality and intelligibility based on standardized evaluation metrics. Moreover, our proposed system can significantly improve the ASR performance with a notable 48.28% relative Character Error Rate (CER) reduction (from 14.50% to 7.50%) under quiet conditions. In an actual noisy environment (sound pressure from 61.7 dBA to 73.9 dBA), our proposed system with a BCM outperforms an ACM, yielding an 84.46% reduction in the relative CER (our proposed system: 9.13% and ACM: 58.75%).
Senior Smart Band:
Modern medicine enables a larger segment of the population to survive beyond 65 years old. Senior care has become one of the single most relevant challenges globally. There are many relevant issues related to monitoring and managing the daily activities of senior citizens. In this study, we will examine the role of advanced activity sensor platform to monitor their daily activity levels. It will assist in understanding the lifestyle of the individual seniors to promote safety and improve the quality of life through measurement of physical strength and independence. We investigated the requirements of wearable technology for the seniors, employing Inertial Measurement Unit (IMU) sensor for senior care. Our proposed system includes a wearable device for each senior, Internet of Thing (IoT) receiver environment, smart alert, machine learning algorithm, and Application Processing Interface (API) with remote Internet access. The primary application of our proposed IMU sensor application includes precise measurement of individual physical activity level.

Acknowledgements iv
中文摘要 v
Abstract Viii
Introduction 1
1.1 Automatic Speech Recognition System 1
1.2 Bone-Conducted Microphone 2
1.3 Recording Environment 3
1.4 ASR Results of ACM and BCM 6
1.5 The Previous Researches Related to BCM 6
1.6 Senior Care 9
1.7 Wearable Technology 10
1.8 Wearable Technology Requirement for Seniors 10
1.8.1 Wireless Protocol 10
1.8.2 Identification 11
1.8.3 Battery Replacement or Power Charge 11
1.8.4 Alert Service 11
1.8.5 Real-Time Location 12
1.8.6 Fall Detection 13
1.8.7 Physical Activity 15
1.8.8 Statistics and Artificial Intelligence 16
1.8.9 AiCare Platform 17
Our Proposed Method 18
2.1 Neural Network and Deep Learning 18
2.2 Restricted Boltzmann Machine 18
2.3 Autoencoder and Denosing Autoencoder 21
2.4 Our Proposed Method for Bone-Conducted Microphone Usage 23
2.4.1 Feature Extraction 23
2.4.2 Training DDAE 25
2.4.3 Testing Phase 27
2.4.4 Speech Intelligibility Index (SII-) based Intelligibility Enhanced (IE) Post-filter 28
2.5 Our Proposed Method for Senior Smart Band Usage 30
2.5.1 Fall Detection 30
2.5.2 Physical Activity 35
Experiment 37
3.1 Experimental Setup for BCM 37
3.2 Experimental Results for BCM 38
3.2.1 Comparison of Spectrogram and Amplitude Envelope 38
3.2.2 Speech Quality and Intelligibility 40
3.2.3 Automatic Speech Recognition 42
3.3 Experimental Results for Senior Smart Band 46
3.3.1 Fall Detection 46
3.3.1.1 Simulated Data 46
3.3.1.2 Day Care Center 48
3.3.2 Physical Activity 50
Discussion 62
4.1 Bone-Conducted Microphone 62
4.2 Senior Smart Band 64
4.2.1 Fall Detection 64
4.2.2 Physical Activity 66
Conclusion 69
5.1 Bone-Conducted Microphone 69
5.2 Senior Smart Band 70
Bibliography 72
Appendices 79
A. Training Data Set 79
B. Testing Data Set 87
Publication List 90

1Administration for Community Living, “Profile of Older Americans,” https://acl.gov/aging-and-disability-in-america/data-and-research/profile-older-americans, 2017.
2ANSI, 1997. American National Standard: Methods for Calculation of the Speech Intelligibility Index: Acoustical Society of America.
3A. Sorvala, E. Alasaarela, H. Sorvoja and R. Myllylä, “A two-threshold fall detection algorithm for reducing false alarms,” 6th International Symposium on Medical Information and Communication Technology (ISMICT), La Jolla, CA, 2012, pp. 1-4.
4B.G. Steele, L. Holt, B. Belza, S. Ferris, S. Lakshminaryan, D.M. Bucher. “Quantitating physical activity in COPD using a triaxial accelerometer,” Chest, 2000, vol. 117, pp. 1359–1367.
5C.A. Werner, “The Older Population: 2010, Census Briefs U.S. Bureau of the Census”, 2010, http://www.census.gov/prod/cen2010/briefs/c2010br-09.pdf.
6Centers for Disease Control and Prevention, “Home and Recreational Safety,” https://www.cdc.gov/homeandrecreationalsafety/falls/adultfalls.html, 2018.
7Chai, L., Du, J. and Wang, Y. N, 2017. Gaussian Density Guided Deep Neural Network for Single-Channel Speech Enhancement. IEEE 27th International Workshop on Machine Learning for Signal Processing (MLSP), Tokyo, pp. 1-6.
8Chen, Z., Watanabe, S., Erdogan, H., and Hershey, J. R., 2015. Speech enhancement and recognition using multi-task learning of long short-term memory recurrent neural networks. Proc. Interspeech, pp. 3274-3278.
9Chris, D., Bo, L., and Rohit P., 2018. Exploring Speech Enhancement with Generative Adversarial Networks for Robust Speech Recognition. International Conference on Acoustics, Speech, and Signal Processing (ICASSP), Canada, SP-L10.2, arxiv 1803.10132v2.
10C.J. Caspersen, K.E. Powell, and G.M. Christenson, “Physical activity, exercise and physical fitness: Definitions and distinctions for health-related research,” Public Health Rep. 1985, vol. 110, pp. 126-131.
11C.M. Cheng, Y.L. Hsu, and C.M. Young, “Development of a Portable System for Physical Activity Assessment in a Home Environment,” Telemedicine Journal and E-Health, 2008, vol. 14, pp. 1044-1056.
12C. Wang et al., “Development of a Fall Detecting System for the Elderly Residents,” 2nd International Conference on Bioinformatics and Biomedical Engineering, Shanghai, 2008, pp. 1359-1362.
13Daniel, M., and Tan, Z.-H., 2017, Conditional generative adversarial networks for speech enhancement and noise-robust speaker verification, Interspeech, pp. 2008-2012.
14Dekens, T., Verhelst, W., Capman, F., and Beaugendre, F., 2010. Improved speech recognition in noisy environments by using a throat microphone for accurate voicing detection. Proc. EUSIPCO, pp. 1978-1982.
15D. E. O''Leary, "Artificial Intelligence and Big Data," in IEEE Intelligent Systems, vol. 28, no. 2, pp. 96-99, March-April 2013. doi: 10.1109/MIS.2013.39
16D. Lim, C. Park, N.H. Kim, S.H. Kim, and Y.S. Yu, “Fall-Detection Algorithm Using 3-Axis Acceleration: Combination with Simple Threshold and Hidden Markov Model,” Hindawi Publishing Corporation Journal of Applied Mathematics, 2014, vol. 2014, pp. 8, Article ID 896030.
17Donahue, C., Li, B., and Prabhavalkar, R., 2018, Exploring speech enhancement with generative adversarial networks for robust speech recognition, Proc. ICASPP.
18E. Casilari, J. A. Santoyo-Ramón, and J. M. Cano-García, “UMAFall: A multisensor dataset for the research on automatic fall detection,” Procedia Computer Science, 2017, vol. 110, pp. 32–39.
19Erdogan, H., Hershey, J. R., Watanabe, S., and Le Roux, J., 2015, Phase-sensitive and recognition-boosted speech separation using deep recurrent neural networks, Proc. ICASSP, pp. 708-712.
20F. Gemperle, C. Kasabach, J. Stivoric, M. Bauer, R. Martin, “Design for wearability,” In Proceedings of the 2nd IEEE Symposium on Wearable Computers, Pittsburg, 1998, pp. 116-122.
21F. Hossain, M. L. Ali, M. Z. Islam and H. Mustafa, “A direction-sensitive fall detection system using single 3D accelerometer and learning classifier,” 2016 International Conference on Medical Engineering, Health Informatics and Technology (MediTec), Dhaka, 2016, pp. 1-6.
22Flanagan, J. L., 2013. Speech Analysis Synthesis and Perception, Ed. 3. Springer Science and Business Media, Germany Berlin.
23Fu, S. W., Hu, T. Y., Tsao, Y., and Lu, X., 2017. Complex spectrogram enhancement by convolutional neural network with multi-metrics learning. Proc. MLSP.
24Fu, S. W., Tsao, Y., and Lu, X., 2016. SNR-aware convolutional neural network modeling for speech enhancement. Proc. Interspeech, pp. 3768-3772.
25Fu, S. W., Wang, T. W., Tsao, Y., Lu, X., and Kawai H., 2018. End-to-end waveform utterance enhancement for direct evaluation metrics optimization by fully convolutional neural networks. IEEE/ACM Transactions on Audio, Speech, and Language, vol. 26, no. 9, pp. 1570-1584.
26Glorot, X., Bordes, A., and Bengio, Y., 2011. Deep sparse rectifier neural networks. Proc. AISTATS, pp. 315-323.
27Google, 2017. Cloud Speech API,” https://cloud.google.com/speech/.
28Graciarena, M., Franco, H., Sonmez, K., and Bratt, H., 2003. Combining standard and throat microphones for robust speech recognition. IEEE Signal Process. Lett., 10(3), 72-74.
29G. Vavoulas, M. Pediaditis, E.G. Spanakis, M. Tsiknakis, “The MobiFall dataset: An initial evaluation of fall detection algorithms using smartphones,” In: Proceedings of the IEEE 13th International Conference on Bioinformatics and Bioengineering (BIBE), Chania, 2013, p. 1–4.
30Haykin, S., 1995. Advances in spectrum analysis and array processing, 3. Pren-tice-Hall, NJ Upper Saddle River.
31Huang, M. W., 2005. Development of Taiwan Mandarin hearing in noise test. Master thesis, Department of speech language pathology and audiology, National Taipei University of Nursing and Health Sciences.
32Hussain, T., Siniscalchi, S. M., Lee, C. C., Wang, S. S., Tsao, Y., and Liao, W. H., 2017, Experimental study on extreme learning machine applications for speech enhancement. IEEE Access, 5, 25542-25554.
33Huang, Z., Li, J., Siniscalchi, S. M., Chen, I. F., Wu, J., and Lee, C. H., 2015. Rapid adaptation for deep neural networks through multi-task learning. Proc. Inter-speech.
34J. Zakir, T. Seymour, and K. Berg, “Big data analytics,” Issues in Information Systems, 2015, vol. 16, iss. 2, pp. 81-90.
35K.M. Diaz, D.J. Krupka, M.J. Chang, J. Peacock, Y. Ma, J. Goldsmith, et al. “Fitbit®: An accurate and reliable device for wireless physical activity tracking,” International Journal of Cardiology, 2015, vol. 185, pp. 138–140.
36Kolbœk, M., Tan, Z. H., and Jensen, J., 2016. Speech enhancement using long short-term memory based recurrent neural networks for noise robust speaker verification. Proc. SLT, pp. 305-311.
37Kuo, H. H., Yu, Y. Y., and Yan, J. J., 2015. The bone conduction microphone parameter measurement architecture and its speech recognition performance analysis. Proc. JIMET, pp. 137-140.
38Lai, Y. H., et al., 2015. Effects of adaptation rate and noise suppression on the intelligibility of compressed-envelope based speech. PloS one, 10, e0133519.
39Lai, Y. H., et al., 2017. A deep denoising autoencoder approach to improving the intelligibility of vocoded speech in cochlear implant simulation. IEEE Transactions on Biomedical Engineering, 64(7), 1568-1578.
40Lai, Y. H., et al., 2018. Deep Learning-Based Noise Reduction Approach to Improve Speech Intelligibility for Cochlear Implant Recipients. Ear and Hearing.
41L. Hutchison, C. Hawes, and L. Williams, “Access to Quality Health Service in Rural Areas—Long-Term Care,” Rural Healthy People 2010: A companion document to healthy people 2010, Vol. 3, The Texas A&M University System Health Science Center, School of Rural Public Health, Southwest Rural Health Research Center, College Station, TX, 2010, pp. 1-28.
42Li, J., Deng, L., Gong, Y., and Haeb-Umbach, R., 2014. An overview of noise-robust automatic speech recognition. IEEE/ACM Transactions on Audio, Speech, and Language Processing, 22(4), pp. 745-777.
43Liu, Z., Zhang, Z., Acero, A., Droppo, J., and Huang, X. D., 2004. Direct filtering for air- and bone-conductive microphones. Proc. MMSP, pp. 363-366.
44Loizou, P. C., 2007. Speech Enhancement: Theory and Practice. CRC Press, Florida Boca Raton.
45Lu, X., Tsao, Y., Matsuda, S., and Hori, C., 2013. Speech enhancement based on deep denoising autoencoder. Proc. Interspeech, pp. 436-440.
46Lu, X., Tsao, Y., Matsuda, S., and Hori, C., 2014. Ensemble modeling of denoising autoencoder for speech spectrum restoration. Proc. Interspeech, pp. 885-889.
47L.Y. Zhu, P. Zhou, A.L. Pan, J. Guo, W. Sun, X. H. Chen, Z. Liu , and L. Wang, “A Survey of Fall Detection Algorithm for Elderly Health Monitoring,” IEEE Fifth International Conference on Big Data and Cloud Computing, Dalian, 2015, pp. 270-274.
48Martens, J., 2010. Deep learning via Hessian-free optimization. Proc. ICML, pp. 735-742.
49Meng, Z., Li, J., Gong, Y., and Juang, B.-H., 2018. Adversarial teacher-student learn-ing for unsupervised domain adaptation. Proc. ICASSP.
50M.E. Rida, F. Liu,Y. Jadi, A.A.A. Algawhari, and A. Askourih, “Indoor Location Position Based on Bluetooth Signal Strength,” 2nd International Conference on Information Science and Control Engineering, Shanghai, 2015, pp. 769-773.
51Mimura, M., Sakai, S., and Kawahara, T., 2017, Cross-domain speech recognition using nonparallel corpora with cycle-consistent adversarial networks. Proc. ASRU.
52Odelowo, B. O., and Anderson, D. V., 2017, Speech enhancement using extreme learning machines. Proc. WASPAA, pp. 200-204.
53P. Pierleoni, A. Belli, L. Palma, M. Pellegrini, L. Pernini and S. Valenti, “A High Reliability Wearable Device for Elderly Fall Detection,” IEEE Sensors Journal, 2015, vol. 15, no. 8, pp. 4544-4553.
54P. Vallabh, R. Malekian, N. Ye and D. C. Bogatinoska, “Fall detection using machine learning algorithms,” 24th International Conference on Software, Telecommunications and Computer Networks (SoftCOM), Split, 2016, pp. 1-9.
55Santiago, P., Antonio, B., and Joan, S., 2017, SEGAN: Speech enhancement genera-tive adversarial network. Interspeech, pp. 3642-3646.
56S.B. Khojasteh, J.R. Villar, C. Chira, V.M. González, and E. de la Cal, “Improving Fall Detection Using an On-Wrist Wearable Accelerometer,” Sensors 2018, 18, 1350, doi:10.3390/s18051350.
57Shimamura, T., Mamiya, J., and Tamiya, T., 2006. Improving bone-conducted speech quality via neural network. Proc. ISSPIT, pp. 628-632.
58Shimamura, T. and Tomikura, T., 2005. Quality improvement of bone-conducted speech. Proc. ECCTD, pp. 1-4.
59Shivakumar, P. G. and Georgiou, P. G., 2016. Perception optimized deep denoising autoencoders for speech enhancement. Proc. Interspeech, pp. 3743-3747.
60S. Kajioka, T. Mori, T. Uchiya, I. Takumi, and H. Matsuo, “Experiment of indoor position presumption based on RSSI of Bluetooth LE beacon,” IEEE 3rd Global Conference on Consumer Electronics (GCCE), Tokyo, 2014, pp. 337-339.
61S. F. Hossain, M. Z. Islam and M. L. Ali, “Real time direction-sensitive fall detection system using accelerometer and learning classifier,” 4th International Conference on Advances in Electrical Engineering (ICAEE), Dhaka, 2017, pp. 99-104.
62S.M. Cheer and A.J. Wagstaff, “Quetiapine. A review of its use in the management of schizophrenia,” CNS Drugs, 2004, vol. 18 iss. 3, pp. 173-199.
63Sun, L., Du, J., Dai, L.-R., and Lee, C.-H., 2017, Multiple-target deep learning for LSTM-RNN based speech enhancement. Proc. HSCMA, pp. 136-140.
64Taal, C. H., Hendriks, R. C., Heusdens, R., and Jensen, J., 2011. An algorithm for intelligibility prediction of time–frequency weighted noisy speech. IEEE Transactions on Audio, Speech, and Language Processing, vol. 19, no. 7, pp. 2125-2136.
65Tajiri, Y., Kameoka, H., and Toda, T., 2017. A noise suppression method for body-conducted soft speech based on non-negative tensor factorization of air- and body-conducted signals. Proc. ICASSP, pp. 4960-4964.
66Thang, T. V., Kimura, K., Unoki, M., and Akagi, M., 2006. A study on restoration of bone-conducted speech with MTF-based and LP-based models. Journal of Signal Processing, pp. 407-417.
67T.R. Mauldin, M.E. Canby, V. Metsis, A.H.H. Ngu, and C.C. Rivera, “SmartFall: A Smartwatch-Based Fall Detection System Using Deep Learning,” Sensors 2018, 18, 3363, doi:10.3390/s18103363.
68U. Lindemann, A. Hock, M. Stuber, W. Keck, and C. Becker, “Evaluation of a fall detector based on accelerometers: a pilot study,” Medical & Biological Engineering & Computing, New York, 2005, vol. 43, pp. 1146–1154.
69van Hoesel, R., et al. , 2005. Amplitude-mapping effects on speech intelligibility with unilateral and bilateral cochlear implants. Ear and Hearing, 26, 381-388.
70Wand, M. and Schmidhuber, J. (2017). Improving speaker-independent lipreading with domain-adversarial training. arXiv preprint arXiv:1708.01565.
71Wang, D., and Chen, J., 2017, Supervised speech separation based on deep learning: an overview. arXiv preprint arXiv:1708.07524.
72Wang, Q., Rao, W., Sun, S., Xie, L., Chng, E. S., and Li, H., 2018, Unsupervised do-main adaptation via domain adversarial training for speaker recognition, Proc. ICASSP.
73Wang, Y. and Wang, D., 2012, Cocktail party processing via structured prediction, Proc. NIPS, pp. 224-232.
74Wang, Y., Narayanan, A., and Wang, D., 2014. On training targets for supervised speech separation. IEEE/ACM Trans. Audio, Speech, Language Process. 22(12), pp. 1849-1858.
75Weninger, F., Erdogan, H., Watanabe, S., et al., 2015, Speech enhancement with LSTM recurrent neural networks and its application to noise-robust ASR. Proc. LVA/ICA, pp. 91-99.
76Wikipedia, 2019. " Autoencoder," https://en.wikipedia.org/wiki/ Autoencoder.
77Wikipedia, 2019. "Restricted Boltzmann Machine," https://en.wikipedia.org/wiki/Restricted_Boltzmann_machine.
78Wikipedia, 2019. "Wearable_Tecnhology," https://en.wikipedia.org/wiki/ Wearable_Tecnhology.
79Xia, B. and Bao, C., 2014. Wiener filtering based speech enhancement with weighted denoising auto-encoder and noise classification. Speech Communication, 60, pp. 13-29.
80Xu, Y., Du, J., Dai, L.R., Lee, C.H., 2014. An experimental study on speech en-hancement based on deep neural networks. IEEE Signal Process. Lett. 21 (1), pp. 65–68.
81Xu, Y., Du, J., Dai, L.-R., and Lee, C.-H., 2015, A regression approach to speech en-hancement based on deep neural networks, IEEE/ACM Trans. Audio, Speech, Language Process. 23, pp. 7-19.
82X. Wu, V. Kumar, J. Ross Quinlan and et al, “Top 10 algorithms in data mining,” Knowledge Information System, 2008, vol. 14, pp 1-37. https://doi.org/10.1007/s10115-007-0114-2.
83Zhang, Z., Liu, Z., Sinclair, M., Acero, A., Deng, L., Droppo, J., Huang, X. D., and Zheng, Y., 2004. Multi-sensory microphones for robust speech detection, en-hancement, and recognition. Proc. ICASSP, pp. 781-784.
84Zheng, Y., Liu, Z., Zhang, Z., Sinclair, M., Droppo, J., Deng, L., Acero, A., and Huang, X. D., 2003. Air- and bone-conductive inte-grated microphones for robust speech detection and enhancement. Proc. ASRU, pp. 249-254.
85Z. Jianyong, L. Haiyong, C. Zili, and L. Zhaohui, “RSSI based Bluetooth low energy indoor positioning,” International Conference on Indoor Positioning and Indoor Navigation (IPIN), Busan, 2014, pp. 526-533.