TD-Expert
Authentication and Evidence Tampering Verification.The Audio Expert speech analysis environment provides several automated tools to search for the traces of editing. These tools are based on the methods developed to prove the authenticity of audio recordings which is an important task on the stage of recognition of audio recording as evidence.
The main idea of authenticity analysis is to check the correspondence of different features of audio to the technical parameters and situational circumstances of the real acoustic act.
Samples Histogram Analysis:
One of these features is the recording format and the samples distribution.
The samples distribution center is quite far from 0.
The histogram of the samples or samples distribution can reveal peculiarities and the individual shape of a particular recording device. Several examples are seen below.
Many samples at 0.
Sharp shape of the samples distribution center.
Linear compression.
Left border of the sample distribution is at -27900 instead of -32768.
The standard procedure includes recording of the test files by the given recorder to prove that the investigated recording was created by it. Histograms of samples distribution comparisons help to prove this.
The evidence recording (blue) and test recording (green) were registered by the same recording device.
Samples distribution must correspond to the file format – for example, compression features. Weather the histogram contains traces of several compression algorithms becomes an important fact that should require an explanation.
Traces of A-law compression intervals in a PCM non-compressed file.
Detailed Waveform Analysis:
Displaying waveform without connection of samples can help to reveal the moments where the compression order was corrupted and give an idea of the way the data was manipulated.
Traces of an A-law compressed interval inside the main PCM non-compressed file. Histogram before (green), inside (red) and after (blue) the interval. This part was injected from another file.
Traces of A-law file samples disorder (at 0:30). Histograms before (green), inside (red) and after (blue) this moment. The signal was processed in sound editing software (weakening).
DC Offset Analysis:
DC offset behavior shows the change of the samples distribution center during the time. The most usual cause of the DC offset jump is the overloading of the signal, but the moments of editing are revealed as well. The idea is based on the fact that there is no reason for the center of the samples distribution to change too quickly. The weighting frame smoothes the braking points but doesn’t affect the general trend.
Sudden change of the DC offset position claims on using several recordings or different parts of a longer recording to frame up the evidence recording.
Digital Start/Stop Search:
Another feature that is quite complicated to detect by visual or aural investigation is the digital start/stop moment that consists of several samples of the same level (the processor of the digital recorder is not stopped though the ADC is switched off already). This effect appears in some of the digital recorder models or particular recorders and can be used as an individual feature in a case. Pushing the start/stop button during the conversation leads to the lack of audio information and loss of authenticity of the recording. That’s why it is so important to detect such moments. The automatic tool executes a search for the coincident samples in the fixed range (to avoid case of overloading detecting) in the whole file. This tool is very effective for long files.
Very short (30 ms) drop-out (22 dB) was found during the automatic search in this file. It was produced by pushing “pause” on the recording device.
Harmonics Phase Analysis:
Narrow band technical harmonics’ phase continuity is good in the case of analog recordings or recordings containing this type of parasite signal. Usually this type of noise sounds as a pure tone, but this investigation method is also good for periodic impulses (“location update” GSM function or any kind of radio impulses). The harmonics energy time dependence is given at the central plot. In case the harmonic doesn’t have phase brakes, a large and short energy increase/decrease must be the explanation.
The harmonics’ brakes.
The harmonic itself has no brakes or jumps (upper illustration) the harmonics’ energy changes together with average level of background noise (middle illustration), but the phase of the harmonic has brakes at 01:49 and 02:06. This picture is very typical for insertion of additional information into the file.
Spectrum Estimation:
Background noise level and structure can be estimated by the spectrogram and highlighted by signed marks of different color. Spectrum view normalization is also available.
Sudden changing of background noise level and structure.
Autocorrelation Analysis:
Dynamic cepstrogramm helps to estimate the continuity of vocal and tone sounds, separate speaker pitch in case of voice cocktail, check the presence of beginnings and ends of the phrases or words.
One speaker (suspect) interrupts another (victim) with a phrase “(Be)lieve me!”. Cepstrogram contains only “Leave me!” (after blue vertical cursor). It is impossible to expect such words from a person that poses a threat to another person.
Average Spectrum Analysis:
Average spectrum reveals signs of resampling or digital processing of initial audio data. Reflection of lower spectrum components from different borders is a trace of resampling or conversion between audio formats.
A sudden jump at 12 kHz shows that the track was previously resampled with 24 kHz.
Reflection of spectrum components at 8, 12, 16 and 20 kHz is the sign of digital processing of the file.
Files’ Header Analysis:
In case the software application leaves signs of audio recording editing in the header of the audio file, Expert reveals those traces and represents them to the user. All other file options are represented as well.
Sonic Sound Forge traces in the files header.
Pricing:
* Quantity discount pricing available.
* 1 year warranty
"Specialized Countermeasures and Intelligence Solutions"
© 2019 - Waypoint Counter Surveillance Inc.
| 
