Hi,
I would like to view and record yes/no response pattern charts to see and analyse my own responses to some questions selected on my own.
Is there avaliable any demo software to let me record voice pattern charts ?
Hi,
While I don't know of any specific "free" CVSA software out there, you can get good results by simply downloading free audio recording software.
Before continuing, here's a little background on myself (w/o giving giving away my secret super-hero identity):
Before working in LE I was an audio engineer (among other things) specializing in psychoacoustics. I am a certified expert in the field of audio analysis, dynamics and physics. After having taken a CVSA myself, I decided to conduct a scientific experiment (to satisfy my curiosity). Although I passed the CVSA, it had nothing to do with telling the truth. It was simply dumb luck that I passed.
Without going into the complete details (it would take many pages to explain the base of comparisions, proceedures, etc.) of the experiment, here is the short version: Using a million dollar audio recording faciltly, equipped with state-of-the-art audio processing, recording and analysing hardware and software, as well as "critical recording" quality condenser microphones costing thousands of dollars, I basically did what I describe below, but in addition I also ran frequency comparisons, THD and S/N ration comparisions to those of what were actual truth "no's" to those of deliberate lie "no's" and "no's" to no questions at all.
After careful study by myself and fellow colleagues, we concluded that "there was no substantial correlation between truth, lie and generic "no's" as to from a consistent pattern of truth/lie beyond that of chance levels."
Anyway, back to how you can conduct your own test.
Just record your responses while in a quiet, controlled environment and then compare the waveform patterns to those of "established" "yes/no, truth/lie" waveforms that are readily available on the internet. You'll see that things don't add up.
In addition, you can simply repeat the word "no" multiple times and compare the WF's to those of supposed "truth" no's and "lie" no's. You'll readliy see that there is no consistency in the patterning. If you simply repeat "no" 50 times and then compare, you'll get about 30 "no's" suggesting that you are lying, even though the word "no" was not in response to any question at all.
With all that said, it's been my intention to submit my findings as a white paper to the AES (Audio Engineering Society), but I have hesitated due to fears that my identity will come into light by the current agency that I work for. Luckily, I never labelled myself as an audio engineer anywhere on my application, as the titles I held lended no hint as to my expertise. Perhaps one day I can share my results with the world.
Anyway, you should find it a fun and insightful experiment.
Good luck!
Thanks OW,
please name which audio recording software worked fine for you.
Could you, as an expert in the field of audio analysis give me a link to some voice patterns recorded by you, as described in the following passage:
"
Using a million dollar audio recording faciltly, equipped with state-of-the-art audio processing, recording and analysing hardware and software, as well as "critical recording" quality condenser microphones costing thousands of dollars, I basically did what I describe below, but in addition I also ran frequency comparisons, THD and S/N ration comparisions to those of what were actual truth "no's" to those of deliberate lie "no's" and "no's" to no questions at all.
After careful study by myself and fellow colleagues, we concluded that "there was no substantial correlation between truth, lie and generic "no's" as to from a consistent pattern of truth/lie beyond that of chance levels."
""
Ok. Just let me to see your voice patterns.
"""
Anyway, back to how you can conduct your own test.
Just record your responses while in a quiet, controlled environment and then compare the waveform patterns to those of "established" "yes/no, truth/lie" waveforms that are readily available on the internet. You'll see that things don't add up.
""""
Ok. Please give me few web links.
"
In addition, you can simply repeat the word "no" multiple times and compare the WF's to those of supposed "truth" no's and "lie" no's. You'll readliy see that there is no consistency in the patterning. If you simply repeat "no" 50 times and then compare, you'll get about 30 "no's" suggesting that you are lying, even though the word "no" was not in response to any question at all.
""""
""""
With all that said, it's been my intention to submit my findings as a white paper to the AES (Audio Engineering Society), but I have hesitated due to fears that my identity will come into light by the current agency that I work for. Luckily, I never labelled myself as an audio engineer anywhere on my application, as the titles I held lended no hint as to my expertise. Perhaps one day I can share my results with the world.
Anyway, you should find it a fun and insightful experiment.
Good luck![/quote]
""""""
Ok. Thanks.
Exactly what I am going to do but a suitable free audio recording software is still misssing.
Have you recorded frequency, energy distribution, voice volume time-shift patterns ?
I just need to make some charts as described in
http://campus.umr.edu/police/cvsa/family.htm
What type they belong in ?
Quote from: OfficerWannabe on Jul 14, 2006, 09:26 PMHi,
Without going into the complete details (it would take many pages to explain the base of comparisions, proceedures, etc.) of the experiment, here is the short version: Using a million dollar audio recording faciltly, equipped with state-of-the-art audio processing, recording and analysing hardware and software, as well as "critical recording" quality condenser microphones costing thousands of dollars, I basically did what I describe below, but in addition I also ran frequency comparisons, THD and S/N ration comparisions to those of what were actual truth "no's" to those of deliberate lie "no's" and "no's" to no questions at all.
....
Anyway, back to how you can conduct your own test.
Just record your responses while in a quiet, controlled environment and then compare the waveform patterns to those of "established" "yes/no, truth/lie" waveforms that are readily available on the internet. You'll see that things don't add up.
...
"THD?" "S/N Ration" [I presume this a typo]
I don't follow your comments at all. THD (Total Harmonic Distortion) and S/N Ratio are terms of art unrelated to the measure of CVSA "microtremors."
Secondly, microtremors (10 Hz range FM modulation of major speech components) aren't visible viewing either the usual time-amplitude waveform or even a time spectrum waveform. The waveforms referenced elsewhere in this thread are clearly neither.
My presumption is that the CVSA is based on stress levels much like the polygraph.
The waveforms appear to be some sort of time vs a summed power distribution of the low frequence modulations CVSA is looking for. While this wouldn't be a particularly difficult signal processing task. I'm not aware of any studio software that provides this.
Quote from: Darius on Jul 15, 2006, 11:18 PMHave you recorded frequency, energy distribution, voice volume time-shift patterns ?
I just need to make some charts as described in
http://campus.umr.edu/police/cvsa/family.htm
What type they belong in ?
Darius: See my comment as to what I suspect those charts in your link represent in an earlier post.
I would suggest using "Protools Free". Check to make sure that your PC's audio card/ hardware setup is compatible. A Sound Blaster should work fine for your needs. You can download PTFree here:
http://www.macmusic.org/software/view.php/lang/en/id/17/
For my testing, I used Protools HD, recorded through a Universal Audio 610 mic preamp, converted with a Rosetta AD/DA converter directly into the Protools HD rig at 192kHz sampling rates. The audio was monitored through a Euphonix CS3000 audio desk.
The mics used were Nuemann U87's.
For the reasons I already mentioned, I don't have any of my results posted online or anywhere else, so I can't supply you with any of my voice patterns right now, sorry.
Here are some exampls of CVSA voice patterns that you can use to compare against your own.
http://www.cvsa1.com/product.php
http://campus.umr.edu/police/cvsa/example.htm
I actually did find some CVSA freeware if you think that will be of more help. I can't attest to the quality of the software, though.
http://www.whatreallyhappened.com/RANCHO/POLITICS/VSA/truthvsa.html
Quote: Have you recorded frequency, energy distribution, voice volume time-shift patterns ?
Yes, frequency ranges, energy (gain) and all other processes inherent to the audio signal were recorded.
We also wrote comaparison software that looked at "micro tremors" and the differences of harmonic and sub-harmonic frequencies between the various "no" responses. However, we expanded the FM modualtion range to 20Hz .It also compared the difference in the "trough and crest" points of the waveforms of the various responses. We recroded "baseline" responses and responses to "lies" and "truths." The software compares the parameters and signal that exceed thresholds were flagged as "lies." In essence, we wrote our own basic version of a CVSA with some difference in the way it displayed the information. The Pro Tool, etc, was simply the method used to aquire the actual audio waveform. We wanted the cleanest, highest quality signal available and we had the luxury of using our own facility.
Please, don't ask me for a copy of the software, as it is not available to the public and I doubt you could you it anyway, as it was written specifically to run in an IRIX workstation.
So again, what I described to you is very basic compared to what actually went into the study. We spent nearly two months conducting the study on our own time after-hours (Geeks tend to take things like this to the max). We did this as a form of entertainment, but we actually took a lot away in the end.
Good luck.
OW, Darius,
Thanks for the additional info, OW. Dennis's work is pretty well documented though it is in Turbo Pascal. No longer a language I work in. I can still read it though. I may give a crack at converting it to MATLAB.
As an aside, I used SoundBlaster to process some audio vinyl signals (for a very different experiment) and it should do fine for this purpose. The better SBs are surprisingly good.
For portable quality audio work, the PCM-D1 provides capture direct to flash in standard linear WAV format.
Quote from: Marty on Jul 17, 2006, 08:41 PMOW, Darius,
Thanks for the additional info, OW. Dennis's work is pretty well documented though it is in Turbo Pascal. No longer a language I work in. I can still read it though. I may give a crack at converting it to MATLAB.
As an aside, I used SoundBlaster to process some audio vinyl signals (for a very different experiment) and it should do fine for this purpose. The better SBs are surprisingly good.
For portable quality audio work, the PCM-D1 provides capture direct to flash in standard linear WAV format.
Hi There Geek Bunnies from the planet X-IR.wav:
I tried the OW in IRIX, but the referencing indicated error messages for Matlab OS. The SN rations were okay with pepsi but unsatisfactory THD was produced down the low end.
I found VSA freeware (use google) and had it written by an IT SOB for the cost of a small house,
or a large circus tent. It produced a linear, singular wave that inverted at anything below 10Hz. Conversely, at >10Hz the linear signal amplified, presenting as two dissimilar but infinitely parallel verticals.
In short, there is nothing available FOC that will present a voice-gram even vaguely similar to that
produced by the Mcquiston-Ford algorithm. No matter what vacuum you live in.