I'm working on VOIP over Tor. Although I haven't yet decided what apps to use, there's an additional concern: voice recognition. I've found Vocoder, which is a vocoder app for Linux. As I understand it, vocoders separate the voice carrier (the tone producer by your vocal cords) from the signal (the modulation added by your tongue, teeth, lips, etc). They use the signal to modulate some other carrier, such as a guitar or other instrument, or just the output of an oscillator. Given that it's the carrier that makes your voice unique, it seems like vocoders might be enough to disguise your voice. But is that really true? Comments, please.
I don't think that this is true. From my knowledge in the voice compression field, the voice signal is obtained by passing what is called the excitation signal (what you called "carrier") through a filter (what you called "signal"). The filter is a model for the whole vocal tract and it is strongly dependent on the speaker (in other words it can identify you). The excitation can be a quasi-periodic signal for voiced sounds and almost white noise for unvoiced sounds. This is also dependent on the speaker (for instance, the period for the quasi-periodic signal is given by the pitch of your voice). So I would say that it is not enough to change the excitation (or carrier as you called it) to conceal your voice. As further proof, a LPC voice compression will keep just the filter parameters and completely replace the excitation with a periodic signal or white noise, and after decoding you will be able to recognize the speaker with a certain precision.
@Nebulus Thanks. I did wonder about that. Are you aware of Linux apps that can alter filter parameters in real time?
Unfortunately I don't know about such apps. But I'm not sure it's very easy to do it properly... Modifying them enough to make the voice unrecognisable but not too much to keep the voice coherent is no easy task.
i guess this could come in handy while calling people you dont trust or something i guess ,im interested as usual , mirimir just keeps on coming out with the craziest ideas i cant resist xD , thou on a sidenote for IRL stuff wouldnt it be best to use something like a softphone like jitsi with zrtp and srtp that should be more than enough to keep any nosy middlemen out of your conversation
gotcha , well if anything comes up ill be the first one to try it out , been thinkin bout that nym remailer buisness as well , still undecided bout the advantages , give me some time , hell it took me almost 4 months to come around on your vpn|tor|vpn setup ,lols
Because I don't trust them enough yet. Leaving a voice sample might be enough to deanonymize me But it's so much easier to discuss things in voice conversations.
I guess I am a little slow here, but who/what is it that you don't trust? And how would the person that you are going to call know that it is you they are talking to when you have disguised your voice.
Privacy and pseudonymity are my favourite hobbies. I also contribute to support human rights. I don't want this advocacy associated with my true identity, because it might be problematic for potential business clients. I don't share other personal information with clients either. They wouldn't, if I had disguised my voice But maybe they might record my voice, and pass it along to someone else with better analytic tools. Perhaps there might be some unfortunate misunderstanding that attracted interest. I don't like unknown unknowns
As a business owner where trust is paramount, I would say if you are doing something that would really upset potential clients, and your business is important to you, I would advise you don't do it. Trying to be clever about it may well bite you in the bottom at the end of the day. Pete
Thank you, that's good coaching. However, I'm considering the possibility of developing a new career based on my hobbies. I've done that twice already, at 10-20 year intervals.
Any way to incorporate a Text to Speech engine? If you are a fast typer, it could be workable. Either software running on your box: http://ubuntuforums.org/showthread.php?t=1979021 ...or a web page like this one: https://www2.research.att.com/~ttsweb/tts/demo.php PD
A comment to follow up on Peter2150's advice... Even if your voiceprint is not traceable, your style of communication may still be unique to you. The way you use time fillers like um, er, uh for example. Your vocabulary and use of idioms as other examples. If the person analyzing the speech patterns knows you or has access to sufficient information about you it could be possible to make a connection. It is a trade off based on: -Motivation and resources of your "adversary" -Amount of effort and resources you are willing to expend -How much it will cost you if you are discovered
Well it sounds like a lot of extra work only because you want to keep your human rights contribution a secret. But since it is your hobbie (the privacy part I mean) I guess you find it fun doing it too. But I'm not sure the person you will talk to finds it as fun as you Unless you're doing something illegal I wouldn't really care if "they" (who ever they are ) would record your voice.
More or less, if people can figure out who I am, my contributions are less valuable. That would be less of an issue if I were deanonymized through some stupid mistake, of course. There's always a balance between workability and worst-case assumptions
Multiband Formant Shifter. Your vocal tract (the source from which the produced sound of your voice originates) determines your unique timbre and formant frequency. If you manipulate the output in this way, then you won't sound like you. I say multiband in conjunction with the frequency spectrum of your voice where you'll split each band around the fundamental frequency making it somewhat dynamic to every pitch and therefore irreversible. If you can program something like that on Linux, you're golden I'm a sound geek
Do you really need to type? Or could you run Dragon Naturally Speaking that somehow hooks to the Text to Speech engine? Might still be susceptible to speech pattern analysis.
I could see where this is useful for someone who is a high profile whistle blower, particularly against a person/organization that has a lot of sophisticated resources it can bring to bear. I don't think "Deep Throat" would be so safe in today's surveillance age.