Voice Analysis

Voice analysis, or voice print comparisons, are requested to help identify an unknown voice on a recording by testing a known voice to the unknown voice.

(In general legal terminology the “known” example is called or referred to as an “exemplar.”)

The processing is done with a computer-based spectrum analysis comparison of key words from the unknown voice recording and an exemplar voice recording.

An exemplar voice recording is an in-person, or over-the-phone recording of a known person, usually alleged as the unknown voice.

On occasion, a pre-existing exemplar sample may be available

 

Our voice comparison process: A Two Step Process

First, Audio-Video Forensic & Restoration Associates will enhance and transcribe the unknown voice recording using our Digital Audio Workstation (DAW).

Then,  in conjunction with Audiomatrix studios, A/VF&RA will record an exemplar sample at our facility, either in person or over-the phone.

Matching words are selected from both the exemplar and unknown samples and are compared in a computer-based spectrum analysis. This process produces what is generally called an “FFT” or Fast Fourier Transform.

The result is a graphic print displaying two lines for each pair of word samples. The plot lines are superimposed in color and show frequency and amplitude content during specific time frames for the two samples. A specified frequency range of human voice measured in Hertz (Hz) is examined in these graphs. Setting this range disregards higher and lower frequency components which are irrelevant for this type of analysis.

The set of graphs are then visually examined and compared.

Here are some examples of these FFT Graphs:

  • Voice print analysis
    Voice comparison "Me"

 

Second, A/VF&R Associates have years of experience listening to audio recordings, transcribing voices and performing voice analysis allowing us to determine with a reasonable degree of certainty if the voices are likely matches, not likely matches or inconclusive from the evidence presented to us. We refer to this part of the process as qualified, aural or listening.

A subjective but informed impression is reached by repeated listening to the comparison samples. Various speech and voice qualities such as timbre/tone, gender, age, inflection, and cadence are compared.

Some factors are obvious such as male or female, young or old, accents in general, and so forth. Other factors we consider include speech or cadence patterns, how an individual groups words into phrases, emphasis or inflections, and so on.

 

In Conclusion

We consider the entire process to be an important tool or aid in the legal process.

Generally, we finish with a written report including the FFT graphs described above.

It is important to note that our findings are based on professional opinion and interpretation of the graphic and aural results.

No test is 100% guaranteed.