Using the Assessments View

Note: This topic is relevant for Forensics.

When presenting legal evidence to the courts, investigators assess the probability that a suspect is the target being searched for. Their informed opinions help the police or prosecuting authority to decide whether prosecution is sensible on the strength of the voice evidence. It may also help the court reach a decision during trial.

An assessment is a statistical analysis of audio segments:

  • Comparison of a questioned sample with voiceprints in a reference population. This determines variations between similar voices (inter-speaker variation).
  • Comparison of a speaker’s voiceprints with each other. This determines variations in the speaker’s voice (intra-speaker variation).
  • Comparison of a questioned sample to a speaker's voiceprints. This determines the point at which the log likelihood ratio (LLR) voice evidence is computed.

Use the Assessment view to compare audio segments. An assessment is a one-to-one comparison of a questioned sample (the speaker is an unknown person) to a suspect speaker voiceprint. To prove that two or more audios are spoken by the same person, or that one person is known by more than one identity, run multiple assessments for each needed comparison.

Each line represents one assessment. Clicking the column headings sort the list by key attributes:

Column

Description

Timestamp/Name

Date the audio segment was uploaded.

Suspect

The speaker’s name.

Reference population

The reference population used for the assessment.

Calibration set

The calibration set used for the assessment.

Reports

Shows status of assessments, reports, and archives.

Results

The system result.

Verdict

The court’s conclusion.

Related topics