THE UNIVERSITY OF BRITISH COLUMBIA Multimodal Design & Technologies Sidney Fels

Fels: Multimodal Design and Technologies 2 Overview Introduction Human I/O and Technology Bringing Modalities Together –intimate and embodied interfaces Summary

Fels: Multimodal Design and Technologies 3 Introduction Virtual Environments –large, medium, and small scale –include people Communicating human experience –information, emotion, environment –people to people –people to machine Engagement of body and mind

Fels: Multimodal Design and Technologies 4 Introduction Techniques needed for: –sensing, encoding, transmitting, storing, indexing, retrieving, compressing, recognizing and synthesizing Interface should disappear

Fels: Multimodal Design and Technologies 5 Introductions Two messages 1. exploit knowledge of human systems for design and research 2. use knowledge of human system to evaluate new technologies/systems

Fels: Multimodal Design and Technologies 6 Human Information Processing Input visual channel - covered already auditory channel - covered already position and motion sensing channel somatic channel, haptic (touch) channel taste and smell channels Output intentional –neuromuscular, movable, verbal non-intentional / biopotentials –galvanic skin response, heart rate, brain, muscle activation

Fels: Multimodal Design and Technologies 7 Human Information Processing Cognition and decisions –tracking –memory –learning –individuals vs. groups

Fels: Multimodal Design and Technologies 8 Driving Trends in Interface Technologies Virtual Reality (VR), Immersive Environments, Augmented Reality (AR) Ubiquitous computing, Intelligent Environments Wearable Computing, Tangible Bits Games, Arts, Interactive Theatre, Interactive Art WWW, Agents, Collaborative work

Fels: Multimodal Design and Technologies 9 Visual Channel: Retina

Fels: Multimodal Design and Technologies 10 Visual Channel: Eye Properties 2 eyes for binocular vision 100,000 fixation points (100deg circular) Types of eye movement (six muscles): –compensatory (must have target) –pursuit (must have target) –Tremor, flick and drift –saccadic (jump from one fixation to another)

Fels: Multimodal Design and Technologies 11 Visual Channel: Movement 8-10Hz gives sensation of motion. –Note relation to attention/other senses 5 ways to make a light move –Real light –Stroboscopic –Induced –Autokinetic –Movement aftereffect

Fels: Multimodal Design and Technologies 12 Visual Channel: Depth & Size Seeing depth: –Binocular vision: Disparity can be used to determine distance –Other cues: overlap, relative size, relative height, atmospheric perspective, texture gradients, parallel line convergence, motion parallax, accommodation and convergence Seeing size: –size constancy

Fels: Multimodal Design and Technologies 13 Visual Displays Main Issues –field-of-view –resolution –update rates, animation –stereopsis –perspective Main Technologies –CRTs, LCDs, projectors –HMDs, CAVEs, Cubby, FishTank VR, VRDs

Fels: Multimodal Design and Technologies 14 Visual Displays: Issues Field of view vs resolution –f.o.v. = total angular deviation that can be seen horizontal 180 o (no eye movement), 270 o (with eye movement) vertical 120 o –need about 8000x8000 pixels for full resolution –tradeoff f.o.v with resolution using lenses –low resolution -> lack of sharpness can see individual pixels o for immersion single & multiperspective

Fels: Multimodal Design and Technologies 15 Visual Displays: Technologies Cathode ray tube (CRT) Liquid Crystal Display (LCD) Head Mounted Displays (HMD) Projectors –CRTS –LCDs Virtual Retinal Display (VRD)

Fels: Multimodal Design and Technologies 16 Stereo systems –shutter glasses/”flicker glasses” –actively alternate LCD panel to switch between right and left –crosstalk is a problem –many different manufacturers Passive –polarized left and right images –red/green glasses Visual Displays: Stereo Technology

Fels: Multimodal Design and Technologies 17 Visual Displays: Stereo Technology Autostereo Systems –No glasses needed –limited viewing position left/right - use tracking to move active area back/front difficult –Fair amount of research but few products Siggraph 2000/1/2, different systems shown –CRL (Kakeya et. Al.) use Fresnel lens plus comp. Cont. polarizing filters. –NYU, Perlin et al., dynamic parallax barrier using LCDs –Actuality Systems: rotating lights –Twister: Kenji Tanaka, U. Tokyo

Fels: Multimodal Design and Technologies 18 Visual Displays: Autostereoscope Parallax systems –thin barrier –backlit LCD (Dimension Systems Inc.) Lenticular Lenses –3D postcards –Philips and others Holographic Optical Elements (HOE) –RealityVision, Trayner and Orr Dual Concave Mirrors –Dimensional Media holography, lasers, spinning synchronized LEDs –Check out exhibit floor this year and evaluate for yourself

Fels: Multimodal Design and Technologies 19 Visual Displays: Virtual Retinal Display (VRD) scan light directly onto retina no need for screens Work done at HITL at U. of Washington (Furness et al.) –Now done by Microvision

Fels: Multimodal Design and Technologies 20 Visual Displays: Techniques & Systems Virtual Retinal Displays (VRD) Fish-tank VR Cubby CAVE HMDs and VR Other systems: –ImmersaDesk –Actuality Systems’ Perspecta (rotating lights)

Fels: Multimodal Design and Technologies 21 Visual Display Technologies examples of designs exploiting human capabilities –Virtual Retinal Display (VRD) –Cubby –CAVE Automatic Virtual Environment application –Iamascope

Fels: Multimodal Design and Technologies 22 Virtual Retinal Delay (VRD) From HIT lab, U. Of Washington (Furness et al., 1991) Microvision continuing work

Fels: Multimodal Design and Technologies 23 Visual Display: Fish Tank VR Stereo image of 3D scene on a workstation –binocular vision perspective projection coupled to head position –movement parallax cues Advantages: –resolution 2 minutes of arc/pixel simulate depth of field (working area is generally known) stable w.r.t. eye movements –off axis eye movement effects are small (since monitor is far away) virtual workspace does not exclude real workspace

Fels: Multimodal Design and Technologies 24 Visual Display: Fish Tank VR Which is most important for 3D effects –Arthur, Booth and Ware, 1992 –stereopsis or head coupling (HC)? Two experiments –experiment 1: subjects decide which is best 3D appearance head coupled + mono (both eyes) better than without HC –experiment 2: tracing tree paths head coupled + stereo best for understanding complex scenes

Fels: Multimodal Design and Technologies 25 Visual Display: Cubby What if we take Fish-tank VR and use 3 orthogonal screens? –Get CUBBY(J.P. Djajadiningrat, Royal College of Art, London) Uses 3 projectors arranged in a corner Uses headtracked monocular perspective

Fels: Multimodal Design and Technologies 26 Cubby (Djajadiningrat and Gribnau, 2000)

Fels: Multimodal Design and Technologies 27 Visual Display: CAVE What if we make the screens wall sized? –CAVE (CAVE Automatic Virtual Environment)

Fels: Multimodal Design and Technologies 28 CAVE Illustration (U. of Illinois, 1992)

Fels: Multimodal Design and Technologies 29 Application: Iamascope (Fels and Mase, 1997)

Fels: Multimodal Design and Technologies 30 Iamascope Overview Video

Fels: Multimodal Design and Technologies 31 Visual Displays: Summary Visual displays that are: –Low-cost, –immersive, –multi-person, –stereo, –head coupled, and –small do not exist. Opportunity for research

Fels: Multimodal Design and Technologies 32 Auditory Channel Senses mechanical vibration of air molecules –sound travels about 350m/s (1260km/hr) Human range is 16Hz to 20,000Hz Pressure waveform causes hair cells to move (23,500 cells) Perceived loudness is approximately logarithmic Perceived sound is highly dependent upon environment –without reverberation unusual effects are noticed

Fels: Multimodal Design and Technologies 33 Auditory Channel Ears can localize sound using: –Intensity difference between ears better for high frequency (>1000Hz) –Time difference between ears better for low frequency (< 1000Hz) difference must be > 0.03sec (0.65s for complete localization) –Frequency profiles ear shape (pinnae) acts as directional filters See Duda, 1996/7 in HCI Course (#3) (note copyright notice in notes)

Fels: Multimodal Design and Technologies 34 Audio: Benefits eyes free rapid detection alerting backgrounding parallel listening acute temporal resolution affective response auditory gestalt formation –trend spotting

Fels: Multimodal Design and Technologies 35 Audio Technologies: Audio Displays Disadvantages –low resolution –limited spatial resolution –lack of absolute values –lack of orthogonality audio parameters not perceived independently –annoyance – interference with speech – not bound by line of sight – absence of persistence – no printout – user limitations

Fels: Multimodal Design and Technologies 36 Audio: types of sound speech vs. non-speech musical vs. non-musical real vs. synthetic –water drops vs. computer beeps

Fels: Multimodal Design and Technologies 37 Audio: Spatialization Difficulty in reproducing directional sounds makes realism difficult in recordings concert halls, stages etc. If pinna convolution could be simulated we could synthesize sound location –also a function of head and shoulder reflections –need Head Related Transfer Function (HRTF) every person is different

Fels: Multimodal Design and Technologies 38 Audio: Spatialization Range –poorest estimation Cues: –Loudness power drops with range (square law) need info. about sound source –Motion parallax head movement changes azimuth –close source -> large change –far source -> small change Excess interaural intensity difference (IID) – head shadow severe for close objects – single ear sounds (I.e. mosquitos) are threatening Ratio of direct to reverberant sound –add reverb to help with range cues

Fels: Multimodal Design and Technologies 39 Audio Technologies: 3D audio Simple techniques for 3D audio –stereo (two-channel) –surround (I.e. Imax) –binaural recordings

Fels: Multimodal Design and Technologies 40 Audio Technologies: Binaural Binaural recordings

Fels: Multimodal Design and Technologies 41 Audio Technologies: Binaural synthesis Impulse response from source to ear drum: –head related impulse response (HRIR) –freq. response is head-related transfer function (HRTF) With HRTF you only need monaural source HRIR measurements: –Knowles Electronics Manikin for Auditory Research (KEMAR)

Fels: Multimodal Design and Technologies 42 Audio Technologies: HRIR Right ear measurement source in horizontal plane about 0.7 s Pinna reflections Shoulder reflection

Fels: Multimodal Design and Technologies 43 Audio: Convolvotron Developed by Crystal River Engineering –Look for Huron Lake system now in exhibition

Fels: Multimodal Design and Technologies 44 Audio: Getting HRTF Where to get HRTF? –standard HRTF poor elevation no standard available yet –set of standard HRTF's improve by tailoring –individualized HRTF takes time but produces good results –model HRTF parameters adjusted for individuals

Fels: Multimodal Design and Technologies 45 Auditory Channel: Summary Least sensitive on median plane (due to symmetry) Perception of multiple pure tones complicated Ears are well adapted for speech Best 3D sound with –personalized HRTFs –difficult to get

Fels: Multimodal Design and Technologies 46 Haptics touch and force feedback rendering haptic models tenets of physical interaction design Dr. Karon MacLean prepared much of this material

Fels: Multimodal Design and Technologies 47 what is haptic force feedback? a small personal robot: applies computer-controlled forces to user’s hand represents a virtual environment acts as both an input and output device: user feels & controls at same time.

Fels: Multimodal Design and Technologies 48 haptic or “force feedback” interfaces human haptics: psychophysics & cognition machine haptics: machine design, sensing, communications computer haptics: stability, modeling, rendering application

Fels: Multimodal Design and Technologies 49 computer haptics: rendering force feedback

Fels: Multimodal Design and Technologies 50 basic procedure for haptic rendering 1.read device position 2.determine which regime we’re in … changes triggered by, e.g. penetrating a virtual object user- or application-triggered events 3.calculate display force based on mode & position send corresponding force command to motors inform other parts of the system of changes

Fels: Multimodal Design and Technologies 51 direct manipulation of a rigid body rigid body: the VE pretends the user is holding directly onto a fictional rigid body through the handle. dynamic system: the VE pretends the user has a more complex relationship with an independent dynamic system (also fictional) rigid body dynamic system

Fels: Multimodal Design and Technologies 52 background: F = ma K = spring force B = viscosity or damping m = mass or inertia x = position v = velocity a = acceleration  motion determined by F hand - F motor virtual mass-spring model

Fels: Multimodal Design and Technologies 53 rendering a virtual mass-spring-damper F mot is applied to user’s hand (effort) user adjusts position in response (flow)  x changes and is re-measured  F mot (x, v, a) is recalculated. first you must differentiate position to get velocity, acceleration (noisy) or, you can measure acceleration from an accelerometer for a smoother estimate.

Fels: Multimodal Design and Technologies 54 tricks: rendering a wall Basic wall:

Fels: Multimodal Design and Technologies 55 1-way damping to increase passivity a pure spring force for a wall may seem “active” (jittery)  add a dissipative term, where B is the damping coefficient only damp when going into the wall

Fels: Multimodal Design and Technologies 56 tricks: visual / aural illusions examples of metacognitive gap visual: never show the point penetrating the surface, even if it is aural: play a crisp contact sound on contact this makes the surface appear stiffer/harder (very robust effect) actual: displayed: BUT: if time offset too great, opposite effect

Fels: Multimodal Design and Technologies 57 tenets of physical interaction design touchable interfaces: what’s so special about touch? what kind of interactions is it good for?

Fels: Multimodal Design and Technologies 58 special qualities of touch bidirectionality: encompasses intention, manipulation, gesture and perception social loading: intentional, socially invasive and committing gesture and expression: convey functional and emotional signals through touching

Fels: Multimodal Design and Technologies 59 motivations for touching we touch intending to - do a task probe an object poke to elicit a reaction fidget to relieve tension communicate a message verify that an action is completed enjoy aesthetic pleasure or comfort connect physically or emotionally to living things.

Fels: Multimodal Design and Technologies 60 we avoid a touch through belief that it would be dirty painful forbidden too intimate inhibitions to touching X X X X

Fels: Multimodal Design and Technologies 61 information available from touching assessments of an object’s dynamic and material properties. verification of engagement and completion continuous monitoring of ongoing activity and gradual doneness. building mental models for invisible parts of a system. judgments of other people.

Fels: Multimodal Design and Technologies 62 when active touching helps: potential benefits reconfigurability: represent environment changes haptically (e.g. # of knob detents) dealing with complexity: offer cues to user options differentiate buttons merge discrete steps into fluid continuous control gesture transport electronic tool use away from desktop

Fels: Multimodal Design and Technologies 63 when active touching helps: potential benefits comfort and aesthetics: pleasant tactility satisfying motion & dynamics ergonomics bidirectional environment coupling muscle memory personalization affect and communication: adds social context and presence to mediated user- user or user-machine connections.

Fels: Multimodal Design and Technologies 64 challenges for haptic interaction design continuous vs. discrete manual control: many tasks require both; transitions are interesting displaying interaction potential: multiplexed functions on one handle embedding haptic interfaces: customized to specific task context tight sensory coupling: attainable through wisely designed software architecture interaction with other systems independent or not?

Fels: Multimodal Design and Technologies 65 physical interfaces summary: design especially good for interfaces that are –complex –sensorially overloaded –require continuous control designed in conjunction with other senses and task in mind spectrum of interaction approaches, from literal rendering to abstract, metaphorical models

Fels: Multimodal Design and Technologies 66 Olfactory: How do we smell? different theories: –i.e. chemical, infrared absorption different perceptual mappings: –small prism –four odours: fragrant, acrid, burnt and caprylic –domain specific: wine, beer, etc. –many, many others... Acuity is great - 10,000 times more sensitive than taste negative adaptation occurs –you get used to the smell

Fels: Multimodal Design and Technologies 67 Olfactory Interfaces Smell camera –Susnick & Raknow, Nature, 2000, 406, Smell display –liquids, gels, microencapsulation Application Research: –Joe Kaye: inStink, Dollars & Scents, Scent Reminder, and more... –Fels, Gauthier, Smith: Interactive Yoga system Smell Display Research and Products: –DigiScents (bankrupt), TriSenx (bankrupt) –DIVEpak (Southwest Research Institute, 1993) –see notes

Fels: Multimodal Design and Technologies 68 Olfaction Interface Challenges clean air input evacuate air clean output air control breathing space –sealed room with air filtration –air control in front of and behind user –sealed pod –tethered mask –tubes into an HMD from pack –built into HMD

Fels: Multimodal Design and Technologies 69 Taste taste buds for –sensations of sour, salty, bitter, sweet and umami –receptor not completely resolved umami receptor (Zucker et al., 2002) –extremely complex and poorly understood Interacts with olfaction

Fels: Multimodal Design and Technologies 70 Volitional Output: Neuromuscular Motor control associated with cerebral cortex –activate muscles volitional and non-volitional –can see in facial expression oral apraxia

Fels: Multimodal Design and Technologies 71 Volitional Output: Movable Controls verbal control/non-verbal control tongue movement breath control facial control gait Hand/Arm/Shoulder motion –Greene and Heckman, 1994, Livingston, 1983, Reynier, 1993, Sturman, 1992

Fels: Multimodal Design and Technologies 72 Range of Motion

Fels: Multimodal Design and Technologies 73 Biopotentials: Examples Galvanic Skin Response (GSR) –Affective Computing, i.e. Galvactinator, Scheirer and Picard Heart Rate (HR) –2 Hearts Musical System, (McCaig and Fels, 2002) Brain activity (EEG) –Brain Computer Interface (BCI) Muscle activity (EMG)

Fels: Multimodal Design and Technologies 74 Biopotential: Heart Response Resting range around 72 pulses/sec –varies from 45 to 90 normally change related to: – mental state –emotional state measure electrical change during beating –up to 2V can occur –electrocardiogram (EKG) –signal processing of EKG is correlated with stress (Rowe, 1998)

Fels: Multimodal Design and Technologies 75 Multitude of input/output channels –all active at once I/O mechanisms usually depend upon –cognitive context –emotional contexts All these channels available to assist people –complement each other Multimodal looks at: –integration –substitution –complement Summary of Human I/O

Fels: Multimodal Design and Technologies 76 Multimodal Design: User centred and non-user centred Intimacy and Embodiment –automatic behaviour –sources of aesthetics

Fels: Multimodal Design and Technologies 77 Intimacy and Embodiment Want interfaces that feel “good” to use Humans and machines intimately linked –degree of intimacy supported may determine success Types of relationships: –human to human –human to machine

Fels: Multimodal Design and Technologies 78 Intimacy Intimacy is a measure of subjective match between the behaviour of an object and the control of that object. –extension of “control intimacy” from electronic musical instruments analysis (Moore, 1997) High intimacy implies: –object feels like an extension of self –satisfaction derives from interacting with object –emotional expression flows requires cognitive effort to prevent

Fels: Multimodal Design and Technologies 79 Intimacy Contributing factors –consistency –responsiveness –usefullness –learnability –functionality –others

Fels: Multimodal Design and Technologies 80 Intimacy: Embodiment vs. Disembodiment Case 1: Object disembodied from Self Case 2: Self embodies Object Case 3: Self disembodied from Object Case 4: Object embodies Self selfobject self object self objectself

Fels: Multimodal Design and Technologies 81 Intimacy and Embodiment Design Examples Many excellent examples out there of interesting designs –Many at Siggraph: Jam-o-Drum (Blaine et al.), Wooden Mirror (Rozin), etc. –Tangible Bits work (Ishii et al.) –Ubicomp (Weiser and more) –Wearable computing (Mann and others) –Art, Entertainment –Virtual reality

Fels: Multimodal Design and Technologies 82 Intimacy and Embodiment Design Examples Consider design of: –Flowfield (Chen and Fels, 2003)

Fels: Multimodal Design and Technologies 83 Iamascope Interactive multimedia artwork (SIGGRAPH’97, Fels, Mase and Reiners) Participants are put inside a large kaleidoscope Participants movements also map to music Run in real-time Immersive, exciting, satisfying, intimate

Fels: Multimodal Design and Technologies 84 FlowField: Semantics of Caress Investigate whole hand interaction techniques for VR –use Tactex MTC Express –collaboration between Fels, Chen and Schiphorst Idea: –allow users direct manipulation of fluid use particle simulation for fluid –aesthetics was important hand manipulation on hard surface mapped to obstructions in the flow field

Fels: Multimodal Design and Technologies 85 FlowField: Semantics of Caress

Fels: Multimodal Design and Technologies 86 Multimodal interfaces need to consider –human information processing –matching interface to task use complementary modes where appropriate –intimacy and embodiment Plenty of research opportunities Summary

Fels: Multimodal Design and Technologies 87

Fels: Multimodal Design and Technologies 88 2 Hearts System Motivation: –create two person instrument –facilitate human-human communication –explore human intimacy Idea: –map two heart beats to sound and image –intimacy between two people function of ability to control each other’s heart Graeme McCaig

Fels: Multimodal Design and Technologies 89 2 Hearts System Implementation in 2 phases –Phase 1: navigation in musical terrain one heart beat to clear change in music two heart beats together push virtual ball around musical terrain 2D musical score –patches of music connected together

Fels: Multimodal Design and Technologies 90 2 Hearts System Phase 2 –visualization of heart beat relationship –created in CAVE Implementation –use auras to represent heart beat information –two perspectives in the CAVE head tracked + stereo graphics –music mapped as well

Fels: Multimodal Design and Technologies 91 2 Hearts System