skip to main | skip to sidebar

Some random thoughts

Pages

  • Home

Monday, October 18, 2010

Text search in audio-video file using subtitle file

Sometimes it happens that when watching movie/video or listening to a long speech/song we are only interested in the part where speaker is talking of some specific context but we have no option to search for that part other than going through the complete audio/video file. For example: A user may be interested only in listening to that part of the Obama's talk where he talks about Foreign policies rather than going through the complete video. If given an option then the user can search for "Foreign policies" as a text in the video and then can jump/seek directly to that particular instance to watch the part.

How to tackle this problem: To search for some specific word/line in an audio/video file either there is need to implement the speech to text algorithm and then search for the text or some complex speech recognition algorithm both which is not reliable to give 100% accurate results. Rather than following these approaches user can search in the subtitles (.srt or may be some other subtitle format) file associated with that audio/video for that word/line and can jump directly to that instance to see that part of the clip. The format of the subtitle file (.srt) is such that it has time stamp in the beginning of each line followed by the line which the speaker is speaking. So if a text is search in that file (like a normal text search in the .txt file) it can give the time stamp corresponding to that particular text field and then the audio-video player can be browsed or can be jumped (seek) to that particular location of interest. If more than one field is return in the search then the player can return all the time stamps of that search and user can select which position he/she wants the video to jump/seek. This feature can be integrated in the existing audio-video players with an ease because already the players know how to read the subtitle (.srt) files and show it on the video.

 Various advantages from this:
  1. It can be integrated in the existing audio-video players with an ease because already the players know how to read the subtitle (.srt) files and show it on the video. 
  2. It can also support multi language search like Chinese, Hindi, Japanese, etc because a simple text search supports multi language search and the subtitles can be made available in many languages. 
  3. To support this feature if subtitle file is being added (coupled) to the existing or new audio/video files then the increase in the file size is not very significant since the file size of the .srt files is in kbs only. 
  4. For almost all the big movies and songs, subtitles file is already available. Hence it would be easy to support search in the already existing files. 
  5. This idea is not specific to mobile phones only. It can be used for any audio-video player whether it is desktop application or mobile one or it can be separate unit of audio-video player or iPods, etc. 
  6. Since it is a simple text search in a files, it will be pretty fast also and easy to implement. 
  7. Also this technology can be used in video searches over the internet since most of the searches on internet are made through text fields only. So, to search a video over the internet which all contains some text of interest, user can search for such words/phrases in subtitles file and get the relevant video.
 etc...
Posted by Ankit Baphna at 11:22 PM 0 comments Email ThisBlogThis!Share to XShare to FacebookShare to Pinterest
Labels: mobile, subtitle search, video search

Avatar based Multifunctional Display in-sync with gadgets


Car in synchronized with wired/wireless gadgets like mobile/iPod/Body Sensors (ECG/BP/Sugar)/ Car remote/GPS and having a fully interactive Avatar based Artificial Intelligence powered Panel to communicate/talk/take actions and give suggestions to the driver and other folks sitting in the car (the human like faces that you see are named as Avatar). Avatar based panel will make drive full of fun and interactive, you can talk with it, give instructions and can do loads of things with just voice controls. It should be running on a multi-lingual artificial intelligence based algorithm so that even your grand-mother can talk with it. There should be options to change the Avatar as well say for example if I am driving to office then Avatar should resemble my secretary and say if I am going to picnic then it should resemble to some cool funky guy. Apart from Avatar my car should be in sync with my gadgets like Mobile, iPod, etc. so that my Avatar can look for meeting and inform secretary if I will be late for that, load the new song which I have loaded recently to iPod and play as soon as I get in the car, call ambulance or inform relatives in case of any emergency, etc.
Here are some of the things (but not limited to) which my car should do-

  1. If my car is in sync with Body Sensor networks attached to my body then my Avatar can suggest me to go to doctor if my blood pressure (or anything) goes wrong or can call ambulance and take control over the car and pull it on side of the road. Or it can show me/give directions to reach to the hospital.
  2. My car should have sensors (mostly camera) just not inside the car but also outside it to read sign boards on road side, hoardings, look out for something which interests me like McDonald’s or something and tell me that the store is 100m away now. Also it can automatically open/close the door for the car owner when he/she come towards the car by scanning the picture through outer camera and when he unlocks the car using remote. Using outer camera’s it can also turn the headlights appropriately on turning points.
  3. Avatar should listen to conversations and take appropriate action. For example, if my wife say's it's hot in here then Avatar should decrease the AC temperature automatically and inform. If my kid says he is feeling boring then it should start a game/quiz automatically and play with him.
  4. If it is in sync with Mobile/GPS and knows the route which the driver is going to follow then based on the traffic conjunction on the route, it can suggest alternate routes (can be done with the help of Google maps).
  5. It can inform my secretary to delay the meeting or can start the Video Conferencing inside the car if it gets the info from my mobile Outlook calendar that meeting is starting in 5mins.
  6. Should adjust the seat heights and distance from steering depending upon the height of the driver - It can do this easily by scanning the picture taken from various cameras inside.
  7. With the camera's placed inside and the conversation going on in the car, Avatar should determine the mood of the passengers and can take appropriate action like playing a romantic song with it's my anniversary and I bring my wife in the car, controls the light and air inside also, etc.
 ...etc..

Overall it will open a whole new world of fun and excitement and comfort and convenience and what not if my car has an Artificial Intelligence powered Avatar talking in it and it's in sync with many gadgets that I have. Possibilities are endless.
Posted by Ankit Baphna at 3:37 AM 0 comments Email ThisBlogThis!Share to XShare to FacebookShare to Pinterest
Labels: avatar, car, mobile
Newer Posts Home

Total Pageviews

Followers

Blog Archive

  • ►  2012 (1)
    • ►  February (1)
  • ►  2011 (2)
    • ►  March (1)
    • ►  January (1)
  • ▼  2010 (2)
    • ▼  October (2)
      • Text search in audio-video file using subtitle file
      • Avatar based Multifunctional Display in-sync with ...

About Me

My photo
Hyderabad, Andhra Pradesh, India
View my complete profile
Powered by Blogger.
 
Copyright (c) 2010 Some random thoughts. Designed for Innovation