5.2.10

Chidambaram and the voice

Home minister P Chidambaram is now getting the hang of it. There could be a one of us types in the 26/11 attacks.

“There was a handler in 26/11 whom we have known for long, or suspected for a long time, could be an Indian. That is something we have known for many, many months now... he goes by the name Abu Jindal, but he is not Abu Jindal, that is not his real name. We cannot get a finger on who he is, unless we get a voice sample. And they (Pakistan) won’t give us a voice sample.”

What? Is there some technical device through which you can figure out an Indian voice? Is it is the accent? Then, don’t forget the questions that were raised about the speech patterns of Kasab and his Pak handlers.

Chidambaram replaced Shivraj Patil as Home Minister following the attacks. If they were aware of this for long, and they asked for voice samples, then how did they put it to the Pakistani establishment? There are several other factors, so why would the voice be the deciding one? If they hinted at something, don’t you think the Pak authorities would go to town about it?

More importantly, are we so dependent on Pakistan to provide us with clues?

Now that PC has brought it out in the open, it would be interesting to watch another drama unfold. Did they keep quiet because of the elections and not to 'hurt' a certain section of society?

Incidentally, unlike our reverend home minster, this blogger had raised this point and mentioned it right here on December 1, 2008, five days after the attack:

I know the authorities are coping with several things right now, but has anyone traced the whereabouts of this Deccan Mujahideen group, whatever it is? I mean, we manage to trace crank callers and spam too. This is important. Because they were misleading our security and intelligence agencies.

And where is the Indian Mujahideen? Just wondering....

If you’d like my exclusive peek into the private debate that took place between Pakistan Interior Minister Rehman Malik and India’s Home Minister P Chidambaram regarding the 26/11 attacks and missed it earlier, then go for it.

14 comments:

  1. "What? Is there some technical device through which you can figure out an Indian voice? Is it is the accent? Then, don’t forget the questions that were raised about the speech patterns of Kasab and his Pak handlers."

    Farzana, yes, in fact, a person's voice has a unique digital signature -- a voice can be broken down into a unique digital fingerprint using what is known as Signal Processing, which is a branch of Electronic Engineering. You may have seen devices in movies where a woman talks into a gadget and it sounds like a man with a deep voice, and anyone can mimic anyone else's voice with a large enough recorded sample of a person's speech.

    If the Pakistanis give India an unmarked set of voices, we can figure out which voice belongs to the handler, even without any external labels on them....science rocks.

    ReplyDelete
  2. Dawood Ibrahim has been living in Karachi under the protection of the Pakistani Army s for decades, and he is just one of a horde of criminals and terrorists from India who are considered "Pakistani assets" by the Pakistani army.

    Our Home Minister knows very well that the Pakistanis will not give us anything of worth, but the Pakistani lie that they are cooperating with India cannot be allowed to gain credence, especially when the pakistani army and its chamchas are wailing and beating their chests to want to talk to India....that would be letting them get away with murder.

    The sleight of hand that they want to try on India again is to open "talks" and at the same time turn on the terrorism tap in India (and talks with Pakistan will be immediately followed by a spurt of terrorist incidents all over India). At the same time, they will pretend to the rest of the world that they are being entirely reasonable, and it is India's fault that the dialogue is suffering (this pitch will be raised especially after a terrorist attack in India, when the Pakistanis will bring in Banis, Jews, Hindus, Martians all into one giant conspiracy theory aimed at destroying Pakistan.

    So it appears to me that the HM has asked the Pakistanis for a very specific item that would prove to India that the Pakistanis are serious -- we can all gauge for ourselves the seriousness, or lack of it, on Pakistan's end for this request In terms of high-school chemistry, a litmus test, if you will.

    ReplyDelete
  3. Al:

    Okay, even with my science-ignorant brain, I know there is this voice recognition. But, how can you figure out who this Abu Jindal is when there is no Abu Jindal? Also, a voice test is not admissible in a court.

    The deflection tactics are going on on both sides, I am afraid, for different reasons.

    The Dawood saga is another one...I have written often enough. Again deflection on both sides.

    Btw, regrading the voice thing, it isn't merely science...I can tell by listening to a voice what the person is upto!

    So, intuition rocks!

    ReplyDelete
  4. And George Bush even wanted to save words with 'Read my lips'. PC could try and get adventurous, although it could be the kiss of the spider woman

    ReplyDelete
  5. FV wrote: "But, how can you figure out who this Abu Jindal is when there is no Abu Jindal?"

    India could have voice samples of this "Abu Jindal" by another name, except Pakistanis cannot afford to have the real identity of AJindal revealed.

    In a sense, let's say you are hiding something in you palms behind your back (and no pockets around to hide that), and I know that is the case and want you to reveal what's in your hand. So I insist on shaking hands with you right then and there. Would you reveal what you intend to hide from me or shake my hand?

    ReplyDelete
  6. FV-ji wrote
    "Btw, regrading the voice thing, it isn't merely science...I can tell by listening to a voice what the person is upto!"

    FV-Jiji, But no one outside your head can use that capability. :-)

    Regarding court admissibility, it is a matter of the level of confidence with which the results can be stated. The larger the sample, the closer to 100% the reliability of the results.

    In logical terms, human speech has a small set of "sounds" that are used to make words, so the sample set should contain as much of this list as possible for confidence, and statistically, that is known to be a certain required size.

    Our human mind is the most sophisticated machine there is, as you point out, and I doubt we will scratch the surface of that machine for a few centuries, unless it happens earlier by serendipity, as it sometimes does.

    ReplyDelete
  7. So for example, if the sample voice is speaking sanskrit, urdu, or hindi (or any language with sounds like ksh, jna, zha, etc) and the test sample is speaking french, this technology's results to match the sample and the test would be less reliable, given the huge difference in basic sounds that exist in one language but not the other.

    ReplyDelete
  8. Even the human brain is susceptible to malfunction -- drink some spiked kool aid and one may accurately recognize voices that are not there in reality.

    V.S. Ramachandran and Oliver Sacks have written about how fascinating the machine between our ears really is from their years in the field, though both of them write almost exclusively about "abnormal" brains that behave in strange ways no one can understand.

    ReplyDelete
  9. Al,
    I would agree with you to the point of Voice matching. In speech recognition, Voice matching (and recognition) has seen fairly robust implementations. Though, these techniques are eaqually fairly susceptible to simple distractions such as voice muffling. Heck, even human beings get fooled by it.
    The point where you started walking on a sticky wicket is where you tried mixing the linguistic aspects (acoustic phonetic model, generically speaking) with the voice matching thing. Generic voice matching (approaches looking at different aspects of Voice spectrum) mostly have nothing to do with what gets spoken. Current state of the art in this technology focuses on extracting information out of the spoken voice. A particularly popular public domain example of this technology is Google voice , with it's - albeit - less than satisfactory precision in transcription.
    Cheers,
    Mahesh.

    ReplyDelete
  10. Mahesh, you are absolutely right about generic voice matching that just breaks down the voice spectrum to form the signature, and as you point out...you have a cold, and the matching will fail.

    However, as FV pointed out earlier, I do not believe that is admissible in court, but I could very well be wrong on this. I only know the laws here locally.


    I think for this stuff to be valid in court (if the courts ever allow that), the burden on the prosecution would be to show that the wave signal matches with high precision between the sample and the test (assuming that the subject's voice has nothing abnormal like a cold in both cases).

    Else, the defense could always claim that two people could exist that have the same voice, which is not provable and has to be assumed to be true.

    ReplyDelete
  11. "Else, the defense could always claim that two people could exist that have the same voice, which is not provable and has to be assumed to be true."

    Mahesh, I went up and read some references, and it looks like none of the above are true in Indian Law and Order today.

    There is no notion of acceptable probabilities of confidence in accepting scientific evidence of different types in the Indian Justice system. This needs to change.

    ReplyDelete
  12. Al,
    The voice recognition technology would be more likely used for increased vigilance and narrowing down on the person of interest from the large group of suspects rather than as forensic evidence. Cheers,
    Mahesh.

    ReplyDelete
  13. http://tinyurl.com/yddfy3r

    More on Abu Jindal -- his Indian identity was known to the HM when he made this statement.

    ReplyDelete
  14. Thanks, Al and Mahesh, for this enlightening discussion.

    The moot point now is that the voice sample is only a technique to get different information, and not valid evidence.

    Al...

    In a sense, let's say you are hiding something in you palms behind your back (and no pockets around to hide that), and I know that is the case and want you to reveal what's in your hand. So I insist on shaking hands with you right then and there. Would you reveal what you intend to hide from me or shake my hand?

    If I am hiding it from you, why would I appear before you in this manner? And if it does so happen, and you force me to shake hands, and the object is revealed, it can be dropped and smashed and no trace left. Or I could hand it to you and then it is in your hands, so it is yours :)

    PS: I do not hide things behind my back, except my back.

    ReplyDelete

Note: only a member of this blog may post a comment.