GMM based language identification system using robust features

Abstract

In this work, we have proposed new feature vectors for spoken language identification (LID) system. The Mel frequency cepstral coefficients (MFCC) and formant frequencies derived using short-time window speech signal. Formant frequencies are extracted from linear prediction (LP) analysis of speech signal. Using these two kind of features of speech signal, new feature vectors are derived using cluster based computation. A GMM based classifier has been designed using these new feature vectors. The language specific apriori knowledge is applied on the recognition output. The experiments are carried out on OGI database and LID recognition performance is improved.

Evolving weighting schemes for the Bag of Visual Words

“…Specifically, we plan to pose the problem as one of learning/optimizing the representation matrix, where other evolutionary algorithms could be used. Also, we are interested on learning term-weighting schemes for other domains, like audio [34], time series [46] or accelerometer data [19], and other scenarios as one-shot recognition [21] and early classification [14].…”

“…The success of the bag of words representation in the natural language processing domain has inspired researchers in computer vision as well, and currently the BoVW is among the most used representations for images and videos [20,43,42,30,35,3,9,50,5,28]. In fact, this formulation has trespassed the image and text boundaries, and it has been used for representing audio [34], time series [46], accelerometer [19] signals, etc. In the computer vision analogy, under the BoVW, an image is represented by a vector indicating the importance of visual words for describing the content of the image.…”