@subins2000 I would even fund it!
@rajeesh
Dude, you stole my words!
@subins2000
Subin, when you start your own company (not necessarily an IT company for that matter), please let me know. I will invest a small amount.
@primejyothi @rajeesh 🥺🥺🥺
@subins2000 @primejyothi Let’s talk business (brainstorm, iow).
Subin, how viable would be this anti-ad filter? Presumably the water mark is not publicly available, can it be scraped? How could the list of w/m can be constantly updated?
When it comes to digital ads, dynamic-ad-injection could be prevented but often that leads to loss of many frames in live streams. Can that be improved?
@rajeesh @subins2000 @primejyothi PS, Cisco has given up on set top boxes a while ago. It acquired NDS at a huge price, killed the STB business and laid off many. Also gave up on SOHO segment and Lynksys and other consumer equipments.
@subinpt @subins2000 @primejyothi Most set-top boxes in Indian market are not-well-known Chinese brands and run Linux, isn’t it?
@subinpt @rajeesh @primejyothi hacking into the set top box can bring legal hassle. From what I enquired they have legal clauses to not tamper with the box.
An external device that act as a TV remote would be perfectly legal I think.
Also, since this audio processing is required, a minimum rasp pi would be a requirement. That'd increase the cost.
@subinpt @rajeesh @primejyothi @jishnu Update: I tried out audio fingerprinting. It was easy. I was able to make a script to identify song from microphone. Like Shazam. So, instead of knowing the hidden watermark, the ad sound is enough.
Now I need to buy a rasppi and run it there plus implement the infrared mute option. The next problem is to see how both mute and unmute after ad can be done.
@subins2000 @subinpt @primejyothi @jishnu I have a Raspi lying around. DM your mailing address and you can have it.
@subins2000 @subinpt @primejyothi @jishnu If/when this is usable, I’d first put it use to mute the voice of Manjrekar’s cricket commentary.
@rajeesh Audio fingerprinting and Voice recognition are two different things. For recognizing ads or music, we can use fingerprinting which is relatively easier. But for recognizing a voice, it's more difficult I believe. I'm not sure if a rasppi can handle it, will probably have to use a cloud service. Saw IBM Watson provides one API for it. നോക്കട്ടെ പറ്റുവോന്ന്.
@subins2000 @rajeesh @primejyothi What if the audio signature is played only in the end of the ad or after a certain time from beginning? To mark a mid point or end to make sure that it was fully played and the viewer hasn't changed the channel?