Effects of Lombard-Style Speech Processing, Presentation Level, and Simulated Room Acoustics on Perceived Speech Clarity, Annoyance, and Level Appropriateness

Abstract The perceptual benefit of Lombard-style speech processing under reverberant and noisy conditions remains unclear, particularly because increasing presentation level may improve perceived speech clarity while simultaneously increasing annoyance. This study examined the effects of Lombard-style processing, sound pressure level (SPL), simulated room-acoustic conditions, and stationary masker spectrum on perceived speech clarity, annoyance, and level appropriateness. Twenty Japanese adults with self-reported normal hearing rated a synthesized Japanese lexical token presented over headphones under combinations of ten speech levels (63–90 dB L Aeq ), four simulated reverberation conditions (reverberation time: 0.01–5.93 s), two stationary maskers (brown and speech-shaped noise), and two speech-processing conditions (normal and Lombard-style). Linear mixed-effects analyses showed that perceived speech clarity increased nonlinearly with SPL and approached saturation at higher levels, whereas annoyance increased at an accelerating rate. Lombard-style processing was associated with higher perceived speech clarity and lower annoyance at the centered SPL under the reference masker condition, although these differences varied with presentation level. Greater nominal surface absorption was associated with higher perceived speech clarity and lower annoyance at the centered SPL. Masker spectrum also modestly modified the level-response relationships. Level appropriateness was positively associated with perceived speech clarity and negatively associated with annoyance, with their interaction indicating that the positive association between clarity and appropriateness weakened as annoyance increased. These findings demonstrate a perceptual trade-off between the diminishing clarity benefit and increasing annoyance cost associated with higher speech-presentation levels, with this relationship further influenced by simulated room-acoustic conditions, Lombard-style processing, and masker spectrum under controlled headphone-listening conditions.