See below from Virage's marketing material. I have used this before and it works. You run a video through and it spits out an xml file with time-stamps and metadata about each time stamp: What is being said, on-screen OCR, speaker changes, face recognition, etc. Simply use this metadata file in conjunction with a good pattern matching based search algorithm (Autonomy's IDOL) and you have a great solution for identifying video clips by similarity of the meta-data. QED

Virage VideoLoggerTM
Index and Automate Rich Media With Unparalleled Ease
The award-winning Virage VideoLogger(TM) is the keystone product in Virage's
product offering, forming a critical component in any Enterprise Rich Media
strategy. Virage VideoLogger is the critical first step in any video production
or distribution strategy, scaling to match any environment with an
unparalleled ease of use. Performing multipletasks in real-time, VideoLogger
cuts production costs and transforms video into a highly accessible asset.
As video is being encoded, VideoLogger's advanced capture and analysis
technology works in real time to automatically create a structured index of
the content. The index enables immediate, accurate search and retrieval
of assets. By making video easy to identify, locate and re-use, VideoLogger
allows pieces to be assembled and repurposed faster and with greater
accuracy than ever before.
Virage VideoLogger is currently enabling efficient Enterprise Rich Media management at the UK Department of Trade and
Industry, Italiane Poste, the Canadian National Archive, Case Western University, Cisco and the Library of Congress to name
a few.
Index Video and Other Time coded Information in Real Time
VideoLogger uses advanced technology to form a conceptual understanding
of an analog or digital video signal and to create a structured video index.
VideoLogger automatically extracts information from the signal including
visual storyboard, closed captioning and teletext. Used in combination with
Virage Media Analysis Plug-Ins, it can additionally recognize faces, voices
and types of sounds in the video, identify on-screen text and numbers, and
convert spoken words to text.
VideoLogger can also extract information from external, time coded sources
such as statistical feeds, slide presentations, EDLs, etc. VideoLogger
allows users to further enrich the video database by adding manual
annotations, such as in/out points, titles, descriptions and much more.
Time-synchronized to each encoded copy and to the original content, the
information-rich database provides searchable, pinpoint access to the video.
Once indexed, VideoLogger can be configured to automatically send an email
message to designated persons as an instant alert to the existence of
specified information.
Create Multiple Encodes and Bitrates, Ingest Multiple Formats
Simultaneously during the indexing process, VideoLogger controls the encoding of multiple formats of source content at multiple
bit rates. For example, you can create multiple copies for browsing or publishing to the Web (e.g. QuickTime®, RealVideo®,
Windows Media®, MPEG-1) and, in parallel, for generating program-quality content (e.g. MPEG-2). In addition to
being able to ingest analog video, VideoLogger also supports real-time ingest of a wide variety of digital
video formats, such as those listed above.
Virage VideoLogger delivers advanced capture and analysis for
real-time indexing, processing and managing of content .
Copyright © 2006 Virage. All rights reserved. Other trademarks are registered trademarks and the properties of their respective owners.
[VIR DS] 10.08.06 Product specifications and features are subject to change without notice. Use of Virage software is under license.
Virage Inc. (USA)
One Market, Spear Tower, 19th Floor,
San Francisco, CA 94105, USA
Tel: +1 415 243 9955
Fax: +1 415 243 9984
Email: info@us.autonomy.com
Virage Inc. (UK)
Cambridge Business Park,
Cowley Rd, Cambridge CB4 0WZ, UK
Tel: +44 (0) 1223 448 000
Fax: +44 (0) 1223 448 001
Email: autonomy@autonomy.com
Other Offices
Virage has additional offices in Boston, New York and
Washington DC, as well as in Amsterdam, Beijing,
Brussels, Hamburg, London, Madrid, Milan, Munich,
Oslo, Paris, Rome, Shanghai, Singapore, Stockholm and
Sydney and Taipei.
www.virage.com
Automate and Streamline More Functions
VideoLogger saves manual annotation time by recognizing user-defined characters (such as chevrons in the closed caption
text) to intelligently and automatically segment video into clips. Also, in- and out-points can be added at any time and the
VideoLogger keeps track of their sequential order, thereby enabling non-linear workflows.
Find Clips Instantly
You can save the VideoLogger index to the Virage Solution Server(TM) or to a variety of video environments, including edit suites
and media asset management systems. Users can then use the corporate intranet or the Internet to quickly search and retrieve
video content, much in the same way they navigate the Web for information. Video can be streamed from any point in the
index and sent straight to the desktop.
Use in Any Environment
A powerful application, Virage VideoLogger's open architecture supports Windows XP and Windows 2000 and makes it
easy to integrate into your existing IT infrastructure. Developers can export the video index in any desired format and include
additional data or text analysis information. This flexibility allows searchable video to be incorporated into a broad range of
custom applications.
Features
  Real-time analog video support for NTSC, PAL and
SECAM standards
  Run multiple and simultaneous instances of
VideoLogger on a single machine
  Real-time digital video support for RealVideo,
WindowsMedia, QuickTime an MPEG formats
  Automatic decoding of text and timecode residing
within MPEG digital video files
  Control one or multiple simultaneous encoding
sessions for RealVideo, WindowsMedia, QuickTime
and MPEG
  Automatic FTP upload for encoded video with e-mail
notification
  VTR control for RS-422 or V-LAN compatible devices
  SMPTE and external timecode source compatibility
  Text extraction of closed captions or teletext
  Intelligent and time-based keyframing with adjustable
keyframe sensitivity
  Media Analysis Plug-Ins:
- Standard or Advanced Speech
  Recognition for multiple languages
- Speaker Identification
- Audio Classification
- AutoClip(TM) Identification
- Face Recognition
- On-Screen Text Recognition
- Story Recognition
- Name Extraction
  Automatic e-mail alerts
  EDL Control for automated logging
  Advanced memory and backup architecture for 24x7
operations
  Data export options for VDF, Virage Solution Server(TM),
Oracle, Informix and ALE
  Live and on-demand user annotations and clip
marking
  User-defined video and clip labels
  Configurable function keys for all operations
  Global find and replace capabilities for any text track
  Comprehensive SDK for wide-range of customizations
and integrations into existing processes
  VideoLogger Service and Encoder software for
Windows
  Multi-byte character support (I18n)
  Localized user-interface (L10n)
  MediaSync plug-in for automatic indexing and
synchronization of PowerPoint presentations
Minimum System Requirements*
  2.4 GHz Windows 2000, 2003 or XP workstation
  1 GB RAM
  20 GB hard drive
  CD-ROM drive
  Osprey-100 video capture card
  2 Available PCI slots
Optional Third-Party System
Components
  Video Servers*
- RealVideo server: RealNetworks
- Apple Darwin Server
- Microsoft Windows Media Server
  RS-422 for selected Sony devices
  V-LAN Controller Express V-LAN Controller VideoMedia
*System configurations vary. Please contact Virage for
more information.