Member of the American College of Forensic Examiners

1. Voice identification
2. Audibility analysis
3. Enhancement
4. Authenticity analysis
5. Sound identification
6. Event sequence analysis
7. Dialogue decoding
8. Other signal analysis
9. Telephone Recording

Voice identification is the comparison of a known voice to an unknown voice as an attempt to determine the identity of the unknown voice, or perhaps as a means to eliminate the known voice as a suspect. This involves t-f-a spectrographic ("voiceprint") and machine/computer analysis of the audio signals. Each person's voice has distinct phonetic characteristics which can be captured, analyzed, separated into phonetic components and compared for identification. The original recording (or high quality copy if the original is not available) and exemplars of the suspect speaking exactly the same words as the unknown voice are required for the examination. The original recording should be of good quality (a high percentage of examination requests are rejected due to poor original recordings), and exemplars should be made using similar equipment, if not perhaps the same equipment, as the original. The analysis consists of the comparison of similarities and differences of elements of speech such as bandwidths, fundamental frequency, prosody, vowel formant trajectory, occlusives, plosives, fricatives, pitch striations, formant energy, breath patterns, nasal resonance, coupling, and any special speech pathology of the speaker.

Voice identification evidence remains controversial and faces passionate opposition in courts of law. Some courts have admitted the evidence, while others have not. The United States Supreme Court has declined to review the voice identification cases brought before it.

Audio Enhancement techniques in the field of Forensic Audio are used to improve listenability and/or intelligibility of a sound source:

Listenability enhancement is a relatively simple procedure applying equalization, compression, and perhaps amplitude increase and limiting to an audio recording which is already intelligible, but needs to be made more listenable (i.e., agreeable to listen to).

Intelligibility enhancement is a more complex and demanding engineering task. Sometimes there is the need to process small sections of the recording with region explicit processing specifications, then reassemble the parts into a sequential mirror image of the original. Procedures involve spectrographic analysis, complex deconvolution filtering, wideband and multiband compression, and amplitude increase and limiting. For the forensic audio engineer, this requires a knowledge of the physics of sound, human auditory perception, phonetics, critical bands, as well as knowledge of the electronics and procedures of professional audio. The results can be spectacular, with previously unintelligible dialogue becoming intelligible and, in some cases, previously unknown audio events being revealed.

Authenticity analysis is used to determine if a recording has simultaneous origins as the acoustic events it represents. Suspicious recordings may have their authenticity verified or negated by physical inspection, aural analysis, tape manufacture date verification, frequency, energy, amplitude and formant structure analysis, magnetic tape development, magnified visual inspection and photographic documentation of the results.

Sound identification is the comparison of an unknown sound to a known sound as an attempt to determine the identity of the unknown sound. Often the original recording device will be needed for such comparison. Sound identification techniques can be critical, for example, in determining whether a barely audible sound is a voice or an acoustic illusion caused by the recording device or an extraneous sound source.

Event sequence analysis is the analysis of the sequence of acoustic events to determine such components as timing, frequency and amplitude. This was a technique used to determine the timing between gun shots for the JFK assassination investigation. Such analysis can also verify or negate the authenticity of a recording. If, for example, on the same tape events between A and B are exactly the same as between F and G regarding timing, amplitude, dialogue, same frequencies, etc., then it is a statistical certainty that the recording has been tampered with. The probability of a naturally occurring exact duplication of sequential acoustic events is infinitesimal.

Dialogue decoding is the process of determining from a recording what has been said. Components of speech have patterns which can be displayed in a t-f-a spectrogram. A skilled individual can use this as an aid in making judgments regarding content of words and phrases. As an example, if the spectrogram displays a broadband burst which has a frequency range of approximately 4500 Hz to 8000 Hz, one may reasonably assume this component to be some sort of sibilant sound such as an /s/ rather than a lower frequency vowel such as an /a/. Voiced (vocal chord movement) and unvoiced (no vocal chord movement) areas are graphically displayed, as well as are plosives, fricatives, occlusives and formant patterns specific to certain vowels, consonants and diphthongs, etc. All of these components aid in the process of decoding.

Strictly speaking, dialogue decoding is not within the realm of the forensic audio engineer. However, during the activity of intelligibility enhancement opinions are formed and an experienced engineer's observations may be very astute, since no one has evaluated the recording more repetitively or critically than the engineer. An individual who may be considered an expert in dialogue decoding is a speech scientist. Such individuals are often found in departments of audiology or communication disorders at universities. Certified court reporters are trained to listen to dialogue critically, often transcribing from court recordings, and their skills may be very useful. Also, a well trained musician is trained to critically evaluate sonic events, and such skills may be very helpful in dialogue decoding.

Other signal analysis is used to evaluate "other" sonic events which have been recorded but which may seem extraneous to the investigation. Such audio events may reveal important information, such as background thumps or conversation, machine rhythms, engine frequency, reverberation, or even gunshot discharge signatures. Computer analysis and screen display/comparison of these "other" signals can reveal information that can be of extreme importance when forming opinions and/or determining causes or circumstances of recorded events.

Telephone Recording: Since the advent of digital voice messaging systems and answering machines users have had no way to transfer messages to another medium for preservation or forensic analysis. CAE offers a message retrieval service. For more information and audio examples CLICK HERE.