Speech Recognition Experiment in 'Natural Quiet' Background Noise

Julien Meyer (1) , Fanny Meunier (2) , Laure Dentel (3)

1 L2C2 - Institut des Sciences cognitives Marc Jeannerod - Laboratoire sur le langage, le cerveau et la cognition
2 BCL, équipe Langage et Cognition
3 LeMondeSiffle - Research Association The World Whistles [Paris]

Listening abilities in humans have developed in rural environments which are the dominant setting for the vast majority of human evolution. Hence, the natural acoustic constraints present in such ecological soundscapes are important to take into account in order to study human speech. Here, we measured the impact of basic properties of a typical 'natural quiet' and non reverberant soundscape on speech recognition. A behavioural experiment was implemented to analyze the intelligibility loss in spoken word lists with variations of Signal-to-Noise Ratio corresponding to different speaker-to-listener distances in a typical low-level natural background noise recorded in a plain dirt open field. To highlight clearly the impact of such noise on recognition in spite of its low level, we contrasted the 'noise + distance' condition with a 'distance only' condition. The recognition performance for vowels and consonants and for different classes of consonants is also analyzed.

Mots clés: Word recognition; Vowel/Consonant recognition; Background noise; Distant Speech

Domaines: Psychologie; Neurosciences; Linguistique; Méthodes et statistiques

Julien Meyer, Fanny Meunier, Laure Dentel. Speech Recognition Experiment in 'Natural Quiet' Background Noise. International Congress of Phonetic Sciences, Aug 2015, Glasgow, United Kingdom. ⟨hal-01860821⟩