Promise for the Best Projects
IEEE PROJECTS 2011-12
2823 5816, 93801 02891, 98411 93224
www.ncct.in, www.ieeeprojects.net, firstname.lastname@example.org
Floor, Nungambakkam High Road, Nungambakkam, Chennai – 600 034Near Ganpat Hotel, Above IOB, Next to ICICI
AN ALGORITHM FOR INTELLIGIBILITY PREDICTION OFTIME-FREQUENCY WEIGHTED NOISY SPEECH
Audio, Speech, and Language Processing, IEEE Transactions on
In the development process of noise-reduction algorithms, an objectivemachine-driven intelligibility measure which shows high correlation withspeech intelligibility is of great interest. Besides reducing time andcosts compared to real listening experiments, an objective intelligibilitymeasure could also help provide answers on how to improve theintelligibility of noisy unprocessed speech.In this paper, a short-time objective intelligibility measure (STOI) ispresented, which shows high correlation with the intelligibility of noisyand time–frequency weighted noisy speech (e.g., resulting from noisereduction) of three different listening experiments.In general, STOI showed better correlation with speech intelligibilitycompared to five other reference objective intelligibility models. Incontrast to other conventional intelligibility models which tend to rely onglobal statistics across entire sentences, STOI is based on shorter timesegments (386 ms).Experiments indeed show that it is beneficial to take segment lengths of this order into account. In addition, a free Matlab implementation isprovided.