Long-Term Average Spectrum (LTAS)
Definition
The averaged spectral energy distribution across a long recording segment, capturing the habitual spectral envelope of a speaker's voice. Useful as a language-independent speaker feature in cross-language cases.
- Measures
- Averaged spectral energy over a long recording segment
- Captures
- A speaker's habitual spectral envelope
- Key advantage
- Language-independent speaker feature
- Use case
- Cross-language speaker comparison
Common questions
Why is LTAS particularly useful in cross-language forensic phonetics cases?+
It reflects physiological and habitual vocal tract characteristics rather than language-specific phonemes, so it stays comparable even when a speaker uses different languages across the two samples.
What can distort an LTAS comparison between two recordings?+
Differences in recording channel, microphone, background noise, or bandwidth between the two samples, all of which alter the measured spectrum independently of the speaker's own voice.
Related terms
- Bharatiya Sakshya Adhiniyam 2023 (BSA 2023)
- India's evidence law replacing the Indian Evidence Act 1872, governing admissibility of expert testimony (Section 39) including forensic phonetic evidence. Does not...
- Code-Switching
- Shifting between different registers or even different languages within an interaction, often as a signal of role or authority. An interviewer who...
- DyViS Corpus
- The Dynamic Variability in Speech corpus (Nolan et al., 2009), a UK English speaker database collected under conditions designed to simulate forensic...
- Formant
- A resonance peak in the vocal tract filter; F1 and F2 together characterise vowel quality (height and frontness), while F3 and higher...
- IAFPA (International Association of Forensic Phonetics and Acoustics)
- The professional organisation for forensic phoneticians, founded in 1991 following the collapse of the voiceprint paradigm. Maintains a code of conduct, annual...
- Lexical Tone
- A tone system in which F0 patterns carry word-meaning distinctions. Mandarin (four tones) and Punjabi (three tones) are the forensically most common...
- Praat
- Open-source acoustic analysis software developed by Paul Boersma and David Weenink at the University of Amsterdam, used by virtually all forensic phoneticians...
- Speaker Profiling
- The use of phonetic and linguistic features to infer a speaker's regional origin, language background, age, sex, or social group from a...
- Voice Onset Time (VOT)
- The interval between the release of a stop consonant closure and the onset of vocal fold vibration for the following vowel. VOT...