Speech Analysis

Journalists and activists frequently seek our expertise to uncover corruption and other corporate or state crimes by matching speech patterns in audio recordings to individuals.  In most cases, these investigations involve well-known public figures whose speeches, interviews, and other such samples are accessible in the open-source.

Our speech analysis encompasses a comprehensive examination of fundamental frequency, harmonics, lexical prosody, intonation, and phonemic patterns among two or more speakers.

Speech analysis serves as an invaluable tool in open-source investigations. However, Earshot maintains a fundamental principle that voices are not substitutes for passports, birth certificates, or fingerprints. Earshot acknowledges that voices are formed in sociality and can change significantly depending on context.

Our work primarily focuses on the analysis of speech signals, and we do not possess the expertise to draw sociolinguistic conclusions, such as determining accent.

In 2012, Earshot's founding director, Lawrence Abu Hamdan, worked with asylum seekers from Somalia and Palestine who had been illegitimately denied asylum solely based on speech analysis. Language Analysis for the Determination of Origin (LADO) is instrumental in assessing the validity of asylum claims by tens of thousands of individuals lacking identity documents in Australia, Belgium, Germany, the Netherlands, New Zealand, Sweden, Norway, Finland, Switzerland and the UK. Abu Hamdan’s comprehensive research into the unscientific and unethical practices of LADO led to the provision of advisory information to a senior Dutch immigration judge and the presentation of expert testimony at deportation hearings in the UK asylum and immigration tribunal.

Case Study: Speech Analysis and Sound Authentication for Camfoot

Earshot investigators were asked by Camfoot (La Référence du Football Camerounais) to investigate whether the voices heard on a recording matched those of Samuel Eto'o in a corruption case allegedly involving him.

A prosodic, formant and harmonic analysis of the phonemes /ɶ/ and /ʌ/ suggests with moderate certainty that voice 1 is the voice of Samuel Eto'o. This analysis revealed matching formants (except for the fundamental formant) and consistent harmonic behaviour, suggesting that both voices belong to speakers with similar physiology. Together with a detailed aural analysis, we concluded that it is likely that voice 1 is the voice of Samuel Eto'o.