Lombard Effect in Speech Production by Cochlear Implant Users: Analysis, Assessment and Implications




Journal Title

Journal ISSN

Volume Title




In daily communication, speakers aim to communicate their message in a manner that is intelligible to listeners. When individuals with normal hearing become aware of reduced auditory feedback due to environmental noise, they likely adopt a different speaking style called the ‘Lombard effect’. The Lombard effect is the tendency of speakers to modify their speech production while speaking in the presence of loud noise. Increased levels of masking noise lead to increase in vocal effort including energy, fundamental frequency, and glottal spectral slope. Lombard speech modification is aimed at providing the listener with increased speech intelligibility in challenging listening environments. The Lombard effect is also known to degrade automatic speech systems such as automatic speech recognition (ASR) and speaker identification (SID). While well-studied for normal hearing listeners and automatic speech systems, the Lombard effect has received little, if any, attention in the field of cochlear implant users. To our knowledge, no study has examined whether cochlear implant users employ the Lombard effect during voice communication. This dissertation provided a comprehensive investigation of the research concerning Lombard effect for cochlear implant users with post-lingual deafness. We investigated the nature of the Lombard effect that was produced by cochlear implant users. A variety of acoustic and phonetic features including voice power, fundamental frequency, glottal spectral tilt, phoneme duration, and formant frequencies were analyzed. Mobile personal audio recordings from continuous single-session audio streams collected over an individual’s daily life were used for these analyses. Prior advancements in this domain include the “Prof-Life-Log”longitudinal study at UT-Dallas. The Lombard effect was observed in the speech production of all cochlear implant users. The results indicate that both suprasegmental (e.g., F0, glottal spectral slope and vocal intensity) and segmental (e.g., F1 for /i/ and /u/)) features were altered in such environments. Along with speech production characteristics, the research also focused on the effect of Lombard speaking style on intelligibility by cochlear implant users. A speech corpus for the perceptual experiments of Lombard speech was formulated with normal hearing speakers. A subjective listening test was performed to provide how cochlear implant users respond to Lombard speech in challenging listening environments. The results indicate that the Lombard speech yielded a significant improvement in intelligibility in both quiet and noisy listening conditions. The specific modification of speech production of cochlear implant users under the Lombard effect may contribute to some degree an intelligible communication in adverse noisy environments. Lastly, a practical implication of Lombard effect for developing a speech enhancement algorithm for cochlear implant users was discussed. A previous proposed framework based on Source Generator theory was employed to perturb neutral speech production based on Lombard effect modification. Data from subjective evaluation demonstrated the effectiveness of the proposed speech enhancement algorithm. The results indicated improvement in intelligibility when providing neutral speech which was modified based on Lombard effect properties via the proposed algorithm. The specific variations due to Lombard effect can be leveraged for new algorithm development and further applications of speech technology to benefit cochlear implant users.



Prosodic analysis (Linguistics), Cochlear implants, Speech, Intelligibility of, Speech perception—Testing, Audiometry, Speech



Copyright ©2017 is held by the author. Digital access to this material is made possible by the Eugene McDermott Library. Further transmission, reproduction or presentation (such as public display or performance) of protected items is prohibited except with permission of the author.