Voice Print spectral analysis, speaker authentication, speaker identification,
and voice enhancement
Spectral analysis of voices

The MuScene Studio specializes in many forms of consumer audio arts, from conventional album production to Surround Sound film mastering.  We have expanded our team of engineers to cover the overlapping zone of spatial acoustics, acoustic engineering, psychoacoustics, and acoustic metrics.  This allows us to offer our clients with a new line of service generically categorized as forensic acoustics or audio forensics.  This technology also underpins the rapidly growing market of voice activated security systems.

With the widespread use of smart phones’ recording function, the availability and importance of digitally recorded audio information has increased drastically over the past few years in litigation-related issues.  We faced fast growing demands for services related to authentication and speaker identification, speech enhancement, and voice characteristics analyses since 2013.  In the second half of 2015, we expand our team of engineers to provide quality acoustic services in pre-jurisdictional case evaluation/preparation, as well as civil litigation cases when various aspects of acoustic specialist advice is looked for. 

The figure above is a typical overlay of two speech spectra of the same word/sentence spoken by two speakers.

Through the coupled analysis of digital signal processing algorithms the acoustic sound waves are decomposed into acoustically meaningful components and made visually spread out in multidimensional space, enabling our forensic engineers to determine whether the two spoken messages originated from the same individual.

Similar working principles are applicable to identifying sounds generated by machineries, marine vessels, and armored vehicle engines that find unique value in military applications.


Examples of visualized voice ingredients.

Digital formats for voice samples

You may submit recorded digital voice files in common acoustic file formats, such as WAV, AIFF, mp3, m4a, wma, flac… for our service.  It is essential for our clients to understand that many acoustic factors of your recorded material affect the service cost that applies to your project.  Voice sampling quality, such as signal-to-noise (S/N) ratio, sound pressure (audible volume), presence of clips and distortions, and environment-borne interferences, all contributes to the amount of efforts our sound engineers have to put into analyzing your data file that determines the cost of your project.

Along with your request for service, please provide MuScene with the following information:

1. Sound file for forensic analysis.
2. A written description of the targeted content or goal/purpose for forensic analysis of item 1 above.
3. Sound file containing the utterance of the subject (person) you wish to establish identity/non-identity with. 
4. Supplementary information that you consider related to the interpretation of your project, for example, a specified time range over which you wish authentication performed, the location, surrounding artifacts, distance from the recording device used or recorded phone conversations, and any audible event came to your attention related to the recorded reference utterance of the subject in item 3.

The more precise and detailed information you provide, the closer our analysis outcome will come to answering your question asked, for a given amount of elaboration. 

The MuScene Studio owns advanced acoustic forensic equipment and experienced acoustic forensics engineers capable of signal recovery and enhancement, noise reduction, and high-resolution analysis of “voice-prints”.  We are confident in providing our clients with the most reliable voice forensic analysis.

Shown here is a visual glimpse of an audio editing session.

Range of our service
I.The MuScene Studio offers services in all aspects of audio forensics, including, but not limited to, the followings:

1. Evidence for post-editing or tampering with recorded acoustic data.
2. Presence/absence of audible evidence ascribable to a specific subject in the recorded audio data.
3. Speaker identification by spectroscopic contrast of recorded voice characteristics of a known speaker against an unknown.
4. Multidimensional visual display of characteristic acoustic information.
5. Grouping of regular/irregular features in fundamental and overtone frequencies of speakers 

2-D visual display of speech utterance.

II.
The MuScene Studio provides forensic analysis services on surveillance recordings, audio recordings of meeting progression.
The followings are some examples of the most frequently requested services we received in the past:

1. Noise reduction 
2. Speech enhancement
3. Audio data recovery
4. Archival digital data backup

3-D High-Definition Spectrogram.

III.
The MuScene Studio provides specified services on recorded conversations or sounds generated in marine vessels, cockpit and/or air traffic controlling sessions ascribable to human speech or machinery backdrop:

1. Sampling of sounds generated by know machineries for comparison against the recorded unknown sounds with the assistance of a full range of acoustic spectrographic tools available.
2. Pin down on a specific model or type of machinery generated sounds among a range of reference candidates.
3. Multidimensional spectroscopic presentation of acoustic signals or voiceprints.
4. Speech tonal analysis and pattern recognition.

Please contact us for additional information.

Below is a reference pricing range for your preliminary arrangement of budget planning. 

US $117~2,169.99 /each sample applies when your audio evidence to be scrutinized is not excessively long with good signal quality, and our standard services cover the following scope: audio signal enhancement, noise reduction/filtration, speech interpretation, speaker authentication/identification, audio material authentication, detection of artifacts and tampering, and other customized requirements.   

US $284~8,335.99 /each sample, if the length of audio evidence to be scrutinized is excessively long or the quality of the data is poor such that extensive digital acoustic signal processing and spectrographic analyses is involved.