Video Search and Voice Recognition
Video search engines are popping up and the technologies being used to index video clips are evolving. Currently, most video is indexed by simple tags and text - no different than descriptive tags for images. But voice recognition is developing, where video search engines will comb video content for recognizable words and phrases to return the most accurate results for your search.
Suranga Chandratillake, Co-Founder of video search engine Blinx was part of a panel recently at SES Chicago. While he, among the rest of the video search engines, is excited about voice recognition, the data is still somewhat raw. Chandratillake cites a 50 - 95% voice recognition rate so far. That's pretty broad, but agreed on by the other panelists and video search engine providers. If you're thinking of video as a viable means of marketing or extending your brand recognition, there are a few things to keep in mind in regards to the voice recognition issues.
- Quality: Simply, the better the quality, the better chances that your video will be indexed.
- Accents: Accents play an important role in the quality of voice recognition. Currently, British and American accents are well recognized, but beyond that there begins to be problems.
- Background Noise: Perhaps the single biggest factor. Even if your audio feed is clean, simple background noise can confuse the system. When preparing to shoot video, choose a quiet room or edit your final product carefully.
Chandratillake also mentioned the next step for video search engines - facial recognition. Think Paris Hilton or her recent cohort Britney Spears.