The listening talker: A review of human and algorithmic context-induced modifications of speech

Abstract

International audienceSpeech output technology is finding widespread application, including in scenarios where intelligibility might be compromised - at least for some listeners - by adverse conditions. Unlike most current algorithms, talkers continually adapt their speech patterns as a response to the immediate context of spoken communication, where the type of interlocutor and the environment are the dominant situational factors influencing speech production. Observations of talker behaviour can motivate the design of more robust speech output algorithms. Starting with a listener-oriented categorisation of possible goals for speech modification, this review article summarises the extensive set of behavioural findings related to human speech modification, identifies which factors appear to be beneficial, and goes on to examine previous computational attempts to improve intelligibility in noise. The review concludes by tabulating 46 speech modifications, many of which have yet to be perceptually or algorithmically evaluated. Consequently, the review provides a roadmap for future work in improving the robustness of speech output

Duration of voicing and silence periods of continuous speech in different acoustic environments

“…A speech rate reduction is mainly obtained by inserting more frequent and longer pauses in the speech stream and, to a lesser extent, by speech segment lengthening. 2 A speaking rate decrease is typical of the so-called "clear speech." 1,3,4 Clear speech can be produced in response to instructions to speak clearly, 5 as in the case of clear, read speech, but also spontaneously in order to adapt to a perturbed communication situation or to help a listener with reduced comprehension capability.…”

“…The type of interlocutor and the environment are the main factors of influence. 1,2 Among the strategies adopted to counteract challenging acoustic conditions, a slower speech rate increases the reception of phonetic information and decreases the cognitive effort of the listener. A speech rate reduction is mainly obtained by inserting more frequent and longer pauses in the speech stream and, to a lesser extent, by speech segment lengthening.…”

Talkers produce more pronounced amplitude modulations when speaking in noise

“…Speech produced in noise generally exhibits increased intensity, slower speech rate, raised F0, and flatter spectral tilt (for an overview, see Cooke et al, 2014a). These and other modifications result in what is collectively known as Lombard speech (Lombard, 1911).…”

LiCoS: A New Linguistically Controlled Sentences Test to Assess Functional Hearing Performance

“…Also, the LiCoS test sentences were deliberately recorded using a normal conversational style and speaking rate instead of a less naturalistic form of speech which is optimized for intelligibility ("hyperspeech") [20,21]. Due to the combination of the enhanced linguistic complexity and normal speech rate of the test items, the new LiCoS test provides a highly realistic impression of the differences in day-to-day hearing functioning in oral communication.…”