Ohio Common Pleas Judges Association Winter Conference

December 8th, 2011

Audio Forensics and Voice Spectrograph Analysis

Basic Audio Terminology

Human Ear Threshold of hearing by frequency chart

 

1. Frequency Response Range of Human Hearing:

  • 20Hz to 20kHz.
  • This means that humans can only hear sounds with-in this range. The chart explains that certain frequencies must have a high enough level in decibels for humans to hear them.

 

2. Human Voice Range:

  • Ranges from approximately 85Hz to 255Hz.
  • Typical Male and Female voice ranges differ:

      *Typical Male Voice – 85Hz to 180Hz

*Typical Female Voice – 165Hz to 255Hz

 

Voice energy chart
Typical telephone response

3. Typical response of recording devices and media:
(The Good, the Bad and the Ugly.)

  • High Quality Digital: High sampling frequency, high bit rate
    • Recording studio quality
  • Medium Quality: Analog, tape-based
    • Micro-cassettes and others
  • Low Quality Digital: Low sampling frequency, low bit rate
    • Hand held, solid state devices
With-in each of these 3 domains there are a number of variables that affect sound quality.
New or nearly new devices typically produce better quality recordings.
A wide range of these media formats are presented as evidence.
As technology improves, better recordings become more accessible.

4. The Fletcher-Munson Effect:

  • Also known as the Equal Loudness Contour.
  • Measures perceived loudness levels of different frequencies over time.
  • The difference in actual frequency and perceived loudness can effect sound quality.

 

Fletcher munson chart 2

Importing Audio Evidence to Computer

analog to digital transfers

Analog formats and digital formats

•Analog tape must be recorded into the computer in real-time using professional reproduction devices.

 

•Digital media files come in many forms. They may be imported directly or may require conversion or re-recording.

 

•Internet may be used for transferring digital media.

 

 

 

In every case, the original is preferred.

 

Limitations of analog and digital recordings

•Device placement:

 

–Proximity to desired target

 

–Environmental factors

 

–Concealed versus open placement

 

•Interference

 

–Tape hiss

 

–Phone noise

 

Background noise

Does all digital audio sound good?

•No. As stated above, there are many formats of digital audio recordings. While intelligibility can be improved through processing, results are limited by the original media.

 

•Digital audio is not always superior to Analog audio!

Principles of audio enhancement

 

Q10eq

Principles of audio enhancement

 

•Peak or center frequency

 

•Notch filters

 

•Shelving

 

•High/low pass filters

 

Dynamic level controls

•Compression

 

•Expansion

 

•Limiting

 

L2
xnois

Broadband Noise reduction controls

•Reduction of continuous noise across the frequency spectrum

 

– Hum, white noise, hiss, rumbles, etc.

 

•Sample of background audio without the voice is analyzed first

 

•Filters reduce unwanted frequencies  proportionally

 

•Many parameters must be carefully adjusted!

 

znoise
pazfreq

The FFT: Fast Fourier Transform

•In lay terms, an audio sample is represented as a graph showing its dynamic levels across the frequency spectrum.

 

•Voice frequency range can be visually examined and compared among samples.
Unknown and Exemplar samples
•Unknown samples are usually evidence and come from many sources.
•Exemplar samples also come from different sources. They can be produced by recorded telephone conversation or in person.

 

Voice print analysis
spectrum analysis of 'We'

Subjective analysis of voice samples

•Does it sound like the same person? Factors include:

 

–Age

 

–Gender

 

–Cadence

 

–Inflection

-Other variables

 

 

Transcription formats

•Prepared utilizing a multidisciplinary approach which combines tools from many areas:

 

•Forensic signal restoration and processing

 

•Auditory perception & critical listening

 

•Forensic phonetics & linguistics


script-example
script-example2

Inaudible speech

•The acronym [IA] stands for inaudible. It can be located in place of words, phrases or entire passages to indicate the existence of utterances that cannot be decoded.
Timeline
•Reference times are used throughout the transcript to refer to the audio CD’s track numbers and running times.

 

•Physical examination of evidence

 

•Audio is loaded into the computer

 

•Listen for recording events such as clicks, changes in environment or tone, and continuity of conversation

 

•Waveform examination
  • Spectral examination
Audio file waveform

Courtroom Presentation of Audio Evidence

Audio Video Forensic Expert Testimony and Deposition
•Open court presentations need proper equipment!

 

•Low tech versus High tech courtroom AV systems, there are no standards.

 

•Will the jury have access to the best possible review of audio materials?