Download as PDF, PPTX
![• Mostly in form of time series
• Accelerometer [& gyroscope]
• Most actions in form of distinct, periodic motion patterns
• Walking, running, sitting,..
• Usual features
• Average, standard deviation
• Time between peaks, FFT energy, Binned distribution
• Correlation between axes
• …
Activity Data from Wearable Sensors
41](https://image.slidesharecdn.com/advancedtechnology202223-230414122348-1c777dc6/85/Advanced-Technology-41-320.jpg)
![• Used in many related application domains
• Video surveillance, sports analysis, …
• Advantages
• Rich information
• Disadvantages
• Highly varied activities in natural environment
• Privacy concerns
• Algorithm complexity
Vision Based Systems
[Cheng and Trivedi,2007]
43](https://image.slidesharecdn.com/advancedtechnology202223-230414122348-1c777dc6/85/Advanced-Technology-43-320.jpg)
![• Taxonomy of methods [Aggarwal & Ryoo 2011]
Algorithms
Human
Activity
Recognition
Single Layered
Approach
Space-time
Approaches
Space-time
Volume
Trajectories
Space-time
Features
Sequential
Approaches
Exemplar-based
State-based
Hierarchal
Approach
Statistical
Syntactic
Description-
based
45](https://image.slidesharecdn.com/advancedtechnology202223-230414122348-1c777dc6/85/Advanced-Technology-45-320.jpg)
![• Problems [Pollack 2003 , Horvitz 2002, 2011]
• When to remind?
• What to remind?
• Avoiding activity conflicts
• Solutions
• Planning & scheduling
• Reinforcement learning
Reminders
55](https://image.slidesharecdn.com/advancedtechnology202223-230414122348-1c777dc6/85/Advanced-Technology-55-320.jpg)
![• Simple reminders
• NeuroPager (1994), MAPS (2005), MemoJog (2005)
• AI-based
• PEAT (1997), Autominder (2003)
Reminders
[Davies 2009]
58](https://image.slidesharecdn.com/advancedtechnology202223-230414122348-1c777dc6/85/Advanced-Technology-58-320.jpg)
![• COACH: Monitoring hand-washing activity and prompting
[Mihailidis 2007, U Toronto]
• Vision
• Detecting current state
• Markov Decision process (MDP)
• Prompting
COACH
60](https://image.slidesharecdn.com/advancedtechnology202223-230414122348-1c777dc6/85/Advanced-Technology-60-320.jpg)
![• Opportunity Knocks (OK): public transit assistance [Patterson
2004]
• iRoute: Learns walking preference of dementia patients
[Hossain 2011]
• Commercial
• GPS shoes
• ComfortZone
Outdoor Stray Prevention
ComfortZone
GPS Shoes
Bracelet for
tracking
patients
61](https://image.slidesharecdn.com/advancedtechnology202223-230414122348-1c777dc6/85/Advanced-Technology-61-320.jpg)
More Related Content
![5G Explained! A High Level Overview [Introduction]](https://cdn.slidesharecdn.com/ss_thumbnails/5gexplainedahighleveloverview-260119165306-cc137a3e-thumbnail.jpg?width=640&height=640&fit=bounds)
Advanced Technology
- 1.
- 2. • 【樂齡科技博覽暨高峰會2020】精華片段及展品介紹 • https://www.youtube.com/watch?v=2fYrqySt5IU •樂齡科技 融入生活 • https://www.youtube.com/watch?v=mVcnudQT7lg • 樂齡科技 租返屋企 https://www.youtube.com/watch?v=VkMHbK6H22Y • 【樂齡科技 帶返屋企】 • https://www.youtube.com/watch?v=mlO-4k4SizA
- 3. • In thebeginning, electronic products were designed by young people to be used by young people. • Video games • Computers introduced into schools • Few advertising efforts were made to interest older adults. • Training opportunities were geared toward younger people.
- 4. What do youthink? 1) Older adults are less interested in learning how to use these technologies. 2) Older adults simply cannot learn how to use these technologies. 3) Older adults are more anxious and have poorer attitudes toward computer use relative to younger adults which ultimately leads to nonuse.
- 5. Breakthrough • Most olderadults have positive attitudes toward the use of computers and other types of electronic technology. • Anxiety level did not seem to affect performance. • Older adults did not seem to be more anxious than younger adults in learning how to use electronic technology. • We also found that attitudes could be modified under certain circumstances. • Longer training periods led to more positive attitudes and better performance in the training sessions led to more positive attitudes. • But the effects were small.
- 6. Breakthrough • We alsosearched to find the optimal training method for teaching computer skills to older adults. • We looked at advanced organizers, modeling, manual, and interactive techniques. • We did not find an optimal training method. However, self- pacing and peer interaction seemed to help.
- 7. • Research hasshown that older adults can acquire memory training techniques and software skills, and also glean information on career development, pre-retirement, and/or health issues using advanced technology, e.g. youtube, facebook, instantgram...... • Nowadays, over 70% of elderly computers owners reported that they have Internet access and 80% said they have accessed it in the past month • At present, it is estimated that about 22% of older adults are surfing the Web.
- 8. • National Instituteon Aging • https://www.nia.nih.gov/health • 醫療及老齡化 • https://www.ourhkfoundation.org.hk/zh-hant/research/aging- society?tid=36
- 9. Why is itimportant? • Scope • 8.5 million seniors require some form of assistive care • 80% of those over 65 are living with at least one chronic disease • Every 69 seconds someone in America develops Alzheimer’s disease • Costs • Alzheimer’s Disease: $18,500-$36,000 • Nursing home care costs: $70,000-80,000 annually • Annual loss to employers: $33 billion due to working family care givers • Caregiver gap • Nurses shortage: 120,000 and 159,300 doctors by 2025 • Understaffed nursing homes: 91% • Family caregivers in US: 31% of households • 70% of caregivers care for someone over age 50 • Data from http://www.hoaloharobotics.com/
- 10. By 2030,1 in 5 Americans will be age 65 or older Average life expectancy 81 years By 2040: Alzheimer related costs will be 2 trillion dollars Year Old Population %
- 11. By 2050,1 in 5 person in the world will be age 60 or older
- 12. • An increasein age-related disease • Rising healthcare costs • Shortage of professionals • Increase in number of individuals unable to live independently • Facilities cannot handle coming “age wave”
- 14. • Normal agerelated challenges • Physical limitations • Balance, reaching, etc. • Perceptual • Vision, hearing • Cognitive • Memory, parallel tasks • Chronic age related diseases • Alzheimer’s Disease (AD) • Cancer, advanced disease
- 15. • They needhelp with daily activities • Activities of Daily Living (ADL) • e.g. Personal grooming • Instrumented Activities of Daily Living (IADL) • e.g. Transportation, cooking • Enhanced Activities of Daily Living (EADL) • e.g. Reading, social engagement • Memory Functions • Health monitoring • Removing the burden from caregiver
- 16. Tools & Infrastructure •Smart homes • Mobile devices • Wearable sensors • Smart fabrics • Assistive robotics ? ? ?
- 17.
- 20. Wearable • Applications • Healthmonitoring • Navigation and stray prevention • Mobile persuasive technologies
- 21. Vital Signs • Respirationsensors • Thermal sensors • Galvanic skin response (GSR) sensors • Cardiac Activity • Pulse oximeter • ECG devices • Doppler radars Movement • Accelerometer • Gyroscope Biochemical • Stress markers (lactate in sweat) • Wound healing (pH and infection markers)
- 22. • Pros. • Anywhere,anytime • Portable • Continuous recordings rather than “snapshot “ • Avoid “white coat” syndrome • Cons. • Anywhere, anytime • Should be worn/carried all the time • Wearing a tag can be regarded as stigma • Privacy concern, 24/7 monitoring
- 23. Assistive Robotics • Helpfulin physical tasks • Communication • People consider them as social entities. • Reducing the need for movement
- 27.
- 30. Think • Issues: • Physicalinterference with movement • Difficulty in removing and placing • Weight • Frequency and difficulty of maintenance • Charging • Cleaning • Social and fashion concerns • Suggestions: • Use common devices to avoid stigmatization • Lightweight • Easy to maintain
- 31. Think • Simple Interface •Limit possibility of error • Avoid cognitive overload • Limit options • keep dialogs linear • Avoid parallel tasks • Consider all stakeholders • Patient, formal onsite/offsite caregivers, informal onsite/offsite caregivers, technical personnel
- 32. Think: Ethics andPrivacy • Ethics • Perfect transparency • Control over the system • Fight laziness • Privacy • Encrypt data • Patient authentication (Owner aware)
- 33. Are they readyto adopt? • Healthy older adults use technology more often* • “Not being perceived as useful” * • Better a known devil than an unknown god • Privacy Concerns • Big brother • Stigmatization
- 34. Smart Home Challenges •Smart homes • Location detection • Privacy/unobtrusiveness vs. accuracy • Difficulty with multiple residents • PIR sensor proximity is important • Reliability • Distinguishing anomalies from normal changes • Become more context aware • Standard protocol
- 35. Wearable & MobileChallenges • Wearable & mobile • Power harvesting • Size • Smart fabrics • Limitations when skin is dry or during intense activity • Still hybrid
- 36. • Holter type •Patches • Body-worn • Smart garments • Garment level • Fabric level • Fiber level Wearable Device Types *A. Dittmar; R. Meffre; F. De Oliveira; C. Gehin; G. Delhomme; , "Wearable Medical Devices Using Textile and Flexible Technologies for Ambulatory Monitoring," Engineering in Medicine and Biology Society, 2005. IEEE-EMBS 2005. 27th Annual International Conference of the , vol., no., pp.7161-7164, 2005 36
- 37. Activity Recognition: “Ambient Sensors” Algorithms& Methods: 37
- 38. What is ActivityRecognition? • The basic building block in many applications • Recognizing user activities from a stream of sensor events … A B C D A C D F … An Activity (Sequence of sensor events) A Sensor Event 38
- 39. • Fine grained(individual movements, especially in vision) • Coarse grained (activity) Activity Resolution Movement: e.g. stretching arm Action: e.g. walking Activity: e.g. preparing meal Complexity Group Activity: e.g. team sports Crowd Activity: e.g. crowd surveillance 39
- 40. • More complexactivities need more sophisticated sensors • Sensor networks of PIR sensors, contact switch sensors, pressure sensors, object sensors, etc. • Approaches • Supervised • Probabilistic • Semi/Unsupervised Activity Recognition PIR PIR Floor Pressure Sensors Object Sensor 40
- 41. • Mostly inform of time series • Accelerometer [& gyroscope] • Most actions in form of distinct, periodic motion patterns • Walking, running, sitting,.. • Usual features • Average, standard deviation • Time between peaks, FFT energy, Binned distribution • Correlation between axes • … Activity Data from Wearable Sensors 41
- 42. Activity Recognition: “Vision” Algorithms &Methods: 42
- 43. • Used inmany related application domains • Video surveillance, sports analysis, … • Advantages • Rich information • Disadvantages • Highly varied activities in natural environment • Privacy concerns • Algorithm complexity Vision Based Systems [Cheng and Trivedi,2007] 43
- 44. • Background subtractedblobs and shapes • isolate the moving parts of a scene by segmenting it into background and foreground • Optical flow • Motion of individual pixels on the image plane • Point Trajectories • Velocity, curvature, etc. • … Vision: Low Level Feature Extraction 44
- 45. • Taxonomy ofmethods [Aggarwal & Ryoo 2011] Algorithms Human Activity Recognition Single Layered Approach Space-time Approaches Space-time Volume Trajectories Space-time Features Sequential Approaches Exemplar-based State-based Hierarchal Approach Statistical Syntactic Description- based 45
- 46. • Suitable forrecognition of gestures & actions • Two different representations • Space-time distribution • Data oriented, spatio-temporal features • Sequence • Semantic oriented, tracking Single Layered 46
- 47. • Space-time approachrepresentation • Volume • Trajectories • Local features Space-time Approaches 2D nonparametric template matching, Bobick & Davis, IEEE Trans. Pattern Anal. Mach. Intel, 2001 47
- 48. • Sequential approach •Exemplar: • Directly build template sequence from training examples • State-based • Build a model such as HMM Sequential Approaches y1 y2 y3 y4 x1 x2 x3 x4 … 48
- 49. Hierarchal Statistical: As states (e.g.HHMM, LHMM, …) Syntactic: As Symbols (e.g. CFG, SCFG, ..) Descriptive: As Logical Relations (MLN …) Hierarchal Approach Robust to Uncertainty Encoding Complex Logic Deep Hierarchy 49
- 50.
- 51. • Different typesof context data • Information from sensors • Activities and their structure • User profile & preferences • Static data (e.g. rooms) Context Information 51
- 52. 1. Key-value models e.g.Context Modeling language (CML) 2. Simple markup schema e.g. HomeML 3. Ontology e.g. SOUPA 4. Uncertain context e.g. Meta-data (e.g. freshness, confidence, resolution) 5. Situation modeling & reasoning e.g. Situation calculus Context Modeling Approaches 52
- 53. Indoor Location Identification MethodDisadvantage Smart floor Physical reconstruction Infrared motion sensors Inaccurate, sensing motion (not presence) Vision Privacy Infrared (active badge) Direct sight Ultrasonic Expensive RFID Range WiFi Interference, inaccurate 53
- 54. • Multiple residents •Active Identification • RFID Badges • Anonymous • Motion models (Wilson 2005, Crandall 2009) Person Identification 54
- 55. • Problems [Pollack2003 , Horvitz 2002, 2011] • When to remind? • What to remind? • Avoiding activity conflicts • Solutions • Planning & scheduling • Reinforcement learning Reminders 55
- 56. Assistive Robotics Challenges •Assistive robotics • Marketing and price • Lack of reliable technology • A robot fully capable of helping with all ADLs • Adaptive robots • More user studies
- 57. Applications Cognitive Orthotics • Reminders● Navigation and stray prevention • Planners Health Monitoring • Continuous Monitoring of Vital Signs ● Sleep Monitoring • ADL Therapy & Rehabilitation • Tele-Health Emergency Detection • Fall Detection • Medical emergency Emotional Wellbeing • Social Connectedness • Facilitating Communication 57
- 58. • Simple reminders •NeuroPager (1994), MAPS (2005), MemoJog (2005) • AI-based • PEAT (1997), Autominder (2003) Reminders [Davies 2009] 58
- 59. • Developed byMartha E. Pollack et al. (U. Of Michigan) • Reminders about daily activities • Plan manager to store daily plans • Resolving potential conflicts • Updating the plan as execution proceeds • Models plans as Disjunctive Temporal Problems • Constraint satisfaction approach • Payoff function Autominder 59
- 60. • COACH: Monitoringhand-washing activity and prompting [Mihailidis 2007, U Toronto] • Vision • Detecting current state • Markov Decision process (MDP) • Prompting COACH 60
- 61. • Opportunity Knocks(OK): public transit assistance [Patterson 2004] • iRoute: Learns walking preference of dementia patients [Hossain 2011] • Commercial • GPS shoes • ComfortZone Outdoor Stray Prevention ComfortZone GPS Shoes Bracelet for tracking patients 61
- 62. • SenseCam • MicrosoftResearch, Cambridge, UK, 2004-2011 • Now commercially available as REVUE Memory Aid 62
- 63. • MedSignals • MD.2 MedicationManagement MedSignals 63
- 64.
- 65. Actual Deployments • Patientswith mild form of dementia • Noninvasive deployment • Prompting systems 65
- 66. Prompting Technology • Context-based •Prompt only if task not initiated • Prompt can be re- issued I’ve done this task I won’t do this task I will do it now I will do it later 66
- 67.
- 68. • Issues: • Physicalinterference with movement • Difficulty in removing and placing • Weight • Frequency and difficulty of maintenance • Charging • Cleaning • Social and fashion concerns • Suggestions: • Use common devices to avoid stigmatization • Lightweight • Easy to maintain Wearable & Mobile Design Issues 68
- 69. • Simple Interface •Limit possibility of error • Avoid cognitive overload • Limit options • keep dialogs linear • Avoid parallel tasks • Consider all stakeholders • Patient, formal onsite/offsite caregivers, informal onsite/offsite caregivers, technical personnel User Interface Design Issues 69
- 70.
- 71. • Healthy olderadults use technology more often* • “Not being perceived as useful” * • Better a known devil than an unknown god • Privacy Concerns • Big brother • Stigmatization Are they ready to adopt? *Heart and Kalderon, Older adults: Are they ready to adopt health-related ICT?, 2011 71
- 72. • Smart homes •Location detection • Privacy/unobtrusiveness vs. accuracy • Difficulty with multiple residents • PIR sensor proximity is important • Reliability • Distinguishing anomalies from normal changes • Become more context aware • Standard protocol Smart Home Challenges 72
- 73. • Wearable &mobile • Power harvesting • Size • Smart fabrics • Limitations when skin is dry or during intense activity • Still hybrid Wearable & Mobile Challenges 73
- 74. • Assistive robotics •Marketing and price • Lack of reliable technology • A robot fully capable of helping with all ADLs • Adaptive robots • More user studies Assistive Robotics Challenges 74
- 75. • Legal, ethical •Telemedicine • Lack of regulations • Which state regulations? Patient’s or Physician? • Who is responsible for malpractice? • Risk of fake physicians • Physician out-of-state competition • Insurance & reimbursement • Patient confidentiality Legal & Ethical Challenges 75
- 76. • Technology • Deviceinteroperability • Legal issues • Patient centric • Integrate all • Robots + smart home + wearable/mobile sensors + e-textile • Technology transfer, go beyond prototype Future 76
- 77. Reference: • Lam WaiMan (2019). Palliative care in Hong Kong – past, present and future. https://www.hkcfp.org.hk/Upload/HK_Practitioner/2019/hkp2019vol41Jun/discussion_p aper.html • Wong LY et al (2020). Quality of Palliative and End-Of-Life Care in Hong Kong: Perspectives of Healthcare Providers. https://www.ncbi.nlm.nih.gov/pmc/articles/PMC7400302/pdf/ijerph-17-05130.pdf • Senthil P Kumar and Anand Jim (2010). Physical Therapy in Palliative Care: From Symptom Control to Quality of Life: A Critical Review https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3012236/ • https://www.ourhkfoundation.org.hk/ • http://www.hoaloharobotics.com/ • Toshiyo Tamura etal (2007). E-Healthcare at an Experimental Welfare Techno House in Japan. https://www.ncbi.nlm.nih.gov/pmc/articles/PMC2666468/