Objective characterization of speech intelligibility under continuous fluctuating masker: Comparing speech transmission index (STI) and speech fluctuation quality index (SFQI)

When a listener is exposed to babble noise in a room or space, the reception of speech is influenced by the amount of temporal fluctuations present in the disturbance which in turn is mediated by the sound tail in the room. In order to investigate these conditions a set of simulated signals having different ratios of single talker and babble was prepared for a range of reverberation times. Two regular rooms with volumes equal to 352m3 and 3072m3 respectively were chosen for simulating the sound tails, and the reverberated signals were characterized by means of their "fluctuation strength". It was shown that the overall room dimensions and hence the impulse response details play a minor role whereas the reverberation time drives the behavior of noise fluctuations. Then the objective characterization of speech intelligibility in the rooms was accomplished for a number conditions by means of STI, which is known to behave critically under fluctuating masker conditions. Also the recently proposed speech fluctuation quality index SFQI, which is based on the comparison of fluctuations in the signal with those in the noise, was calculated. This latter indicator effectively separates the different noise fluctuation groups. Finally, recordings in real environments (bar, restaurant, market, bus) were accomplished and a first comparison with simulated conditions was discussed.