Forensic Audio‎ > ‎

OTExpert

«OTExpert» is a powerful Forensic Audio Examination Set that provides audio expert wide spectrum of solutions and reflects the contemporary demands of voice identification and authenticity analysis methods.
In 2003 software application complex «OTExpert» was tested by Scientific Council of Russian Ministry of Justice Forensic Examination Centre and then recommended for usage in all Ministry’ forensic audio labs.
As the producer is highly interested in feedback and the program code is conducted on the contemporary state-of-art basis the wide distribution of «OTExpert» brought a lot of enhancements and new functions due to professional users’ demands. Accessibility of the system architecture and high interface usability provide fast and accurate approach to the signal and speech analysis.


OTExpert


Wide spectrum of functions consists of six major blocks:

Recording and playback

All audio formats and digital settings are supported, input/output level control, clipping notification. Simple editing functions. Loop playback, consequent playback of chosen fragments from two different files, pseudo stereo mode, tempo correction.

Visualization

Oscillogram, spectrogram, LPC-spectrogram, and cepstrogram representation of the signal. Fast switch between modes. Additional information shown over main layer (energy, pitch, formants, user marks and notes). Fast signal navigation. Two clicks to the function settings (frame type and size, spectrum normalization). Fast calculations and information management. Signals synchronization. Rich color presets pallet. Full image export with user presets.

Automatic speaker characteristics extraction

Pitch and formant traces extractor, manual correction options via visual control. Calculation, comparison, and storing of pitch and formants statistics. MS Word export.

Text editor

Color highlighting of speakers’ utterances. Convenient speech segmentation. Text-to-audio binding and fast text navigation. Automatic search of matching words and phrases for further comparison analysis.

Filtering

High-pass, low-pass, band-pass, inverse filtering. FIR, IIR filtering. Easy settings access.

Authenticity analysis

Samples distribution analysis. Technical harmonics’ phase continuity investigation. Files’ header analysis.

Rich and precise functional environment of interrelated procedures implemented via user-friendly interface provide fast and comfortable speech and audio analysis on the high professional level.
Here are some examples of exported images:

Segmentation of the recording and textual transcribing


Identification

Pic. 1. Pitch distributions comparison.

Black regions on dynamic cepstrograms (frequencies – vertical axis, time - horisontal) reflect the behaviour of speakers’ pitch:
Pic. 2. Comparison of prosodic constructions, used by speaker in two recordings.

Pic. 3. Comparison of prosodic constructions, used by speaker in two recordings.

Pic. 4. Comparison of prosodic constructions, used by speaker in two recordings.

Dynamic spectrograms reflect articulators’ behavior. Position of spectral maxima (i.e. formants) correspond to different phonemes, specifying its quality and reflecting the individual features of vocal tract.
Therefore the match of momentum spectra corresponds to match of speakers.

 
Pic. 5. Comparison of [a]-like phoneme pronounced by speaker in two different recordings F1=490Hz, F2=1630Hz, F3=2350Hz, F4=3130Hz
 
Pic. 6. Comparison of [e]-like phoneme pronounced by speaker in two different recordings. F1=440Hz, F2=1460Hz, F3=2330Hz, F4=3200Hz

Pic. 7. Comparison of [o]-like phoneme pronounced by speaker in two different recordings. F1=440Hz, F2=1280Hz, F3=2230Hz, F4=3120Hz 

Pic. 8. Comparison of [i]-like phoneme pronounced by speaker in two different recordings. F1=330Hz, F2=1880Hz, F3=2470Hz, F4=3210Hz 
Pic. 9. Comparison of [a]-like phoneme pronounced by speaker in two different recordings. F1=480Hz, F2=1380Hz, F3=2270Hz, F4=3350Hz
Pic. 10. Comparison of [o]-like phoneme pronounced by speaker in two different recordings. F1=420Hz, F2=1400Hz, F3=2230Hz, F4=3250Hz
Pic. 11. Comparison of [e]-like phoneme pronounced by speaker in two different recordings. F1=400Hz, F2=1900Hz, F3=2420Hz, F4=3450Hz
 


Spectra of consonants can also match:
Pic. 12. Spectra of consonant [č].

Pic. 13. Spectra of consonant [š’].

Authenticity analysis

Brakes of harmonics’ phase:
Pic. 14. Small brakes and smoothed changing of power-supply (50Hz) harmonics’ phase.

Pic. 15. Brakes of 76.56Hz harmonics’ phase.

Pic. 16. Insertion of 16-bit interval into A-law compressed file.

Pic. 17. Additional samples levels on a short interval of a compressed file.

Pic.18. DC Bias shift.

Highly professional experts (www.forenex.biz) provide training courses on the identification task and authenticity analysis using «OTExpert» complex.