Semantic HIFI WP 8 : HIFI System

1
Semantic HIFIWP 8 HIFI System

• September 21st , 2004
• Koen De Vroede

2
Agenda

• Definition of the HiFi system
• Recap various scenarios presented in January
• Feasibility studies carried out by NSCE
• Conclusions
• User requirements specification
• Online survey system
• Summary of findings
• Proposal WP 8 as part of deliverable D 1.2.1

3
Recap various scenarios Jan 2004

• HIFI pc. Good example is Vaio W2.
• Remote controlled pc (UI on remote control or TV)
• Network media receiver (Sony roomlink,
  Soundblaster wireless)
• High end HIFI component (no PC involved)
• HIFI component PC CPU server
• Feedback from kick-off meeting
• Strong preference for a real HIFI system (not
  scenario 1)
• Java is required by many partners
• Interest for advanced remote control

4
Feasibility studies carried out by NSCE

• Candidate platforms for a prototype HiFi system
• Prototype PVR set top box platform running linux
• PS2 linux platform ( PS2 linux dev kit)
• Sony Net Juke NAS A1 (launched 20/5/2004)
• Evaluation based on various criteria
• General hardware resources (CPU, memory, storage,
  peripheral interfaces)
• Audio capabilities (hardware software which
  drivers available)
• Linux distribution (which release, ease of
  upgrading etc.)
• Support for Java?
• Support for IPV6 support for UPnP
• Platform availability (hardware, software)

5
First selection

• Sony Net Juke most promising candidate BUT not
  available in Europe (not even inside Sony) gt
  excluded
• PS2 dropped because of
• Limited RAM (32 MB)
• Default linux kernel is outdated, 2.2.1 and very
  difficult to upgrade and no active linux PS2
  community
• Prototype PVR platform used for investigations
  related to Java functionality
• Selected Kaffe VM over Suns reference
  implementation because Suns VM
• Requires support for X-Windows/Motif for UI
• No support for linux on MIPS
• Requires ALSA audio driver
• Kaffe VM
• Supports MIPS
• Supports various graphics solutions (framebuffer,
  directFB, QT)
• Also supports use of OSS driver ESD daemon

6
Further investigations

• Port of Kaffe VM to PVR prototype platform
• Basic port (interpreter mode) support for JIT
  version remained unstable
• Difficult to get graphics support working on
  target investigation of various possibilities
  (framebuffer, QT, directFB) gt opted for directFB
  with success
• Addition of audio support judged to time
  consuming as no standard linux driver (OSS or
  ALSA) was available (only linux DVB audio API).
• Study of possibilities to split UI backend of
  application (scenarios 2 and 3)
• Preferably opt for UPnP or Java RMI

7
Conclusions

• Opting for Sony proprietary hardware/software
  platform adds serious complexity to the project
  for which no budget is available
• Opting for a distributed architecture (especially
  for all basic functionalities) adds complexity to
  the project where the overall distributed
  architecture (WP 1.2 and WP 7) is already complex
  and still requires further clarification gt drop!
• BUT
• If the goal remains to develop a prototype high
  end HIFI component (so not a PC)
• Windows as software platform is NOT an option
• It is important to apply realistic hardware
  constraints for such a device
• Opt for a platform that is easily accessible
  (hardware, software, dev env) for all partners
• Opt for a platform that represents a realistic
  evolution by 2006 of a high-end HIFI system
  (extrapolate from Sony Net Juke for example)

8
Proposal for HIFI platform

• Hardware platform
• Main unit
• mini-PC based on VIA EPIA ME6000 fanless
  motherboard OR VIA EPIA Nehemiah M10000 1GHz
  variant
• CD/DVD drive (no burner)
• 128 MB RAM
• internal hard disk (40 GB) (can be increased but
  affects price)
• onboard audio (with SPDIF), graphics (TV-out),
  networking, USB (used for WiFi dongle), wireless
  microphone (WIFI interference?)
• One PCI bus (optionally better soundcard MIDI
  support)
• LCD/Plasma TV
• IR remote control, wireless keyboard, PDA with
  WiFi
• PC in home network (optional) authoring, content
  management,
• Software platform
• Linux distribution based on 2.6 kernel,
  postgreSQL, Java with use of directFB
• Gcc tools, Eclipse for Java development

9
Online survey system
10
Survey system data

• Goal was to gather user feedback within 20-30
  minutes
• Studied all use cases, combined similar ones and
  presented them as 28 questions, split in 3
  categories
• Organizing listening
• Sharing
• Creating
• Most questions asked the user to rate the
  proposed functionality (interested not
  interested dont know)
• 16 additional questions studied user profile
  (music preferences, HIFI equipment, etc.)
• System deployed online inside Sony Brussels
  operations (targeting 600 people). We obtained
  115 responses.

11
Summary user profile

• Majority of people aged between 25 - 40
• Nearly everyone owns or has frequent access to a
  PC and 76 uses is for music playback
• WMP and winamp most popular, I-tunes low
• Majority of PC users use Windows, even the linux
  users (27)
• 30 no music collection, bulk has a collection gt
  1 GB only 10 gt 10GB
• Nearly everyone owns at least one HIFI system
• 75 owns a portable device, bulk still CD players
  but solid state players rising
• Nearly 50 owns a home cinema system, and this
  will increase to 70 by 2006
• 50 plans to purchase a flat TV by 2006

12
Summary findings

• Categorize answers in 5 groups
• Very desirable feature 90 people interested 15
  not
• Desirable feature around 70 people interested,
  30 not
• Rather desirable interested gt non-interested
  greater amount of undecided
• Tie in interested non-interested big amount
  of undecided
• Not desired clear majority not interested
• Organizing listening
• Very desirableBasic Playlist support display of
  metadata search on attributes
• Desirable playlists according to attributes
  playlists with transitions find by similarity
  display lyrics in sync
• Rather desirable editing of attributes save
  search criteria Query by Humming play video
• Tie-in navigate within track audio summary

13
Summary findings (2)

• Sharing
• Very desirable none
• Desirable sharing of audio sharing of media
  content related to track support for user groups
• Rather desirable sharing of attributes
• Tie-in none
• Not desirable none
• Creating
• Very desirablenone
• Desirable none
• Rather desirable none
• Tie-in volume balance lead/accompaniment
  virtual remixing track modifications
  (normalize,)
• Not desirable interactive scenarios user
  performs systems transform input tempo
  modification

14
WP 8 project proposal
8.1 HIFI platform (M2) 8.2 Classical HIFI app
limited browsing (M3) (UI audio framework,
IR) 8.3 Indexing browsing (month 25) 8.4 Home
network ShF sharing (incl WP1.2 middleware
data mgmt (m21) 8.5 HIFI remote controller (PDA
with WIFI no advanced UI on PDA) (M4) 8.6
Rendering performing (M7) 8.7 Final overall
integration documentation
 
15
WP 8 project proposal details

• 8.3 Indexing browsing
• Adding audio to local database reorganizing
  local music collection
• Audio summary (IRCAM)
• High level metadata computation iconic/textual
  rep (IRCAM)
• High level descriptors computation (CSL)
• Browsing by attributes finding by similarity
  (CSL)
• Playlists (authoring tool embedded constraint
  solver) (CSL)
• Editing metadata (CSL)
• Query by Humming (FhG)
• 8.7 Rendering performing (VST, MIDI??)
• Virtual audio mixing (BGU)
• Browsing by lyrics (BGU)
• Virtual acoustics (IRCAM)
• On the fly mixing (webcam?) (IRCAM, CSL)
• Audio visual rendering (IRCAM)
• Conducting tempo modification (UPF)
• Smart karaoke (dynamic adaptation user input)
  (IRCAM, UPF)