This thesis presents a new approach to the visualization of sound for deaf assistance that simultaneously illustrates important dynamic sound properties and the recognized sound icons in an easy readable view. .In order to visualize general sounds efficiently, the MFCC sound features was utilized to represent robust discriminant properties of the sound. The problem of visualizing MFCC vector that has 39 dimension was simplified by visualizing one-dimensional value, which is the result of comparing one reference MFCC vector with the input MFCC vector only. New similarity measure for MFCC feature vectors comparison was proposed that outperforms existing local similarity measures due to their problem of one to one attribute value calculation that leaded to incorrect similarity decisions. Classification of input sound was performed and attached to the visualizing system to make the system more usable for users. Each time frame of sound is put under K-NN classification algorithm to detect short sound events. In addition, every one second the input sound is buffered and forwarded to Dynamic Time Warping (DTW) classification algorithm which is designed for dynamic time series classification. Both classifiers works in the same time and deliver their classification results to the visualization model. The application of the system was implemented using Java programming language to work on smartphones that run Android OS, so many considerations related to the complexity of algorithms is taken into account. The system was implemented to utilize the capabilities of the smartphones GPU to guarantee the smoothness and fastness of the rendering. The system design was built based on interviews with five deaf persons taking into account their preferred visualizing system. In addition to that, the same deaf persons tested the system and the evaluation of the system is carried out based on their interaction with the system. Our approach yields more accessible illustrations of sound and more suitable for casual and little expert users.
[1]
John H. L. Hansen,et al.
Unsupervised audio stream segmentation and clustering via the Bayesian information criterion
,
2000,
INTERSPEECH.
[2]
Jonathan Foote,et al.
Visualizing music and audio using self-similarity
,
1999,
MULTIMEDIA '99.
[3]
Tara Matthews,et al.
Visualizing non-speech sounds for the deaf
,
2005,
Assets '05.
[4]
Elena Deza,et al.
Encyclopedia of Distances
,
2014
.
[5]
Malcolm Slaney,et al.
Construction and evaluation of a robust multifeature speech/music discriminator
,
1997,
1997 IEEE International Conference on Acoustics, Speech, and Signal Processing.
[6]
M. A. Siegler,et al.
Automatic Segmentation, Classification and Clustering of Broadcast News Audio
,
1997
.
[7]
Xin Zhang.
Audio segmentation, classification and visualization
,
2009
.
[8]
Andrey Temko,et al.
Acoustic Event Detection and Classification
,
2007,
Computers in the Human Interaction Loop.
[9]
Biing-Hwang Juang,et al.
Fundamentals of speech recognition
,
1993,
Prentice Hall signal processing series.