Babble speech: acoustic and perceptual variability
نویسندگان
چکیده
The presence of babble noise is one of the most difficult environments to sustain speech system performance. This study focuses on acoustic and perceptual analyses of babble. The acoustic variability of babble is analyzed as a function of the number of speakers in babble. The concept of acoustic volume is proposed and it is shown that the acoustic volume reduces as the number of speakers in babble increase. This framework is evaluated using over 40 hours of simulated babble from SWITCHBOARD corpus. It is observed that the acoustic volume does not change significantly when there are 4 or more speakers in babble. It is also seen that with an increase in the number of speakers in babble, there is an uneven spread of data in the acoustic space due to mixing of multi-speaker content. This study provides a framework to better understand the babble environment, enabling us to improve the formulation of reliable algorithms for robust speech systems.
منابع مشابه
Variability in noise-masked consonant identification
Speech communication commonly occurs in the presence of noise. Perceptual errors in the perception of noise-masked speech vary as a function of noise type (e.g., white noise, speech-shaped noise, multi-talker babble), listener characteristics (e.g., listeners with hearing loss, non-native listeners), and target stimulus properties (e.g., native language of the talker, casual vs clear speech). T...
متن کاملPerception of English minimal pairs in noise by Japanese listeners: does clear speech for L2 listeners help?
This study investigates the effect of different types of clear speech on cross-linguistic perception. In particular, it examines whether the words produced to help L2 interlocutors are perceived more easily by them than the words produced for L1 interlocutors to counteract speech babble noise. English keywords in these conditions are extracted from the LUCID corpus, and they are presented to Ja...
متن کاملPersian Phone Recognition Using Acoustic Landmarks and Neural Network-based variability compensation methods
Speech recognition is a subfield of artificial intelligence that develops technologies to convert speech utterance into transcription. So far, various methods such as hidden Markov models and artificial neural networks have been used to develop speech recognition systems. In most of these systems, the speech signal frames are processed uniformly, while the information is not evenly distributed ...
متن کاملThe effect of bilateral subthalamic nucleus deep brain stimulation (STN-DBS) on the acoustic and prosodic features in patients with Parkinson’s disease: A study protocol for the first trial on Iranian patients
Background: The effect of subthalamic nucleus deep brain stimulation (STN-DBS) on the voice features in Parkinson’s disease (PD) is controversial. No study has evaluated the voice features of PD underwent STN-DBS by the acoustic, perceptual, and patient-based assessments comprehensively. Furthermore, there is no study to investigate prosodic features before and after DBS in PD. The curren...
متن کاملAcoustic correlates of intelligibility enhancements in clearly produced fricatives
Two experiments investigated whether and how clear speech production enhances intelligibility of English fricatives for normal-hearing listeners and listeners with simulated hearing impairment. Babble thresholds were measured for minimal pair distinctions. Clear speech benefited both groups overall; however, for impaired listeners, the clear speech effect held only for sibilant pairs. Correlati...
متن کامل