Jatiya Kabi

Continuous Bangla Speech Segmentation using Short-term Speech Features Extraction Approaches

This paper presents simple and novel feature extraction approaches for segmenting continuous Bangla speech sentences into words/sub-words. These methods are based on two simple speech features, namely the time-domain features and the frequency-domain features. The time-domain features, such as short-time signal energy, short-time average zero crossing rate and the frequency-domain features, such as spectral centroid and spectral flux features are extracted in this research work. After the feature sequences are extracted, a simple dynamic thresholding criterion is applied in order to detect the word boundaries and label the entire speech sentence into a sequence of words/sub-words. All the algorithms used in this research are implemented in Matlab and the implemented automatic speech segmentation system achieved segmentation accuracy of 96%.

Human Face Classification using Genetic Algorithm

The paper presents a precise scheme for the development of a human face classification system based human emotion using the genetic algorithm (GA). The main focus is to detect the human face and its facial features and classify the human face based on emotion, but not the interest of face recognition. This research proposed to combine the genetic algorithm and neural network (GANN) for classification approach. There are two way for combining genetic algorithm and neural networks, such as supportive approach and collaborative approach. This research proposed the supportive approach to developing an emotion-based classification system. The proposed system received frontal face image of human as input pattern and detected face and its facial feature regions, such as, mouth (or lip), nose, and eyes. By the analysis of human face, it is seen that most of the emotional changes of the face occurs on eyes and lip. Therefore, two facial feature regions (such as lip and eyes) have been used for emotion-based classification. The GA has been used to optimize the facial features and finally the neural network has been used to classify facial features. To justify the effectiveness of the system, several images were tested. The achievement of this research is higher accuracy rate (about 96.42%) for human frontal face classification based on emotion.

Blocking Black Area Method for Speech Segmentation

Speech segmentation is an important sub problem of automatic speech recognition. This research is concerned with the development of a continuous speech segmentation system using Bangla Language. This paper presents a dynamic thresholding algorithm to segment the continuous Bngla speech sentences into words/sub-words. The research uses Otsu’s method for dynamic thresholding and introduces a new approach, named blocking black area method to identify the voiced regions of the continuous speech in speech segmentation. The developed system has been justified with continuously spoken several Bangla sentences. To test the performance of the system, 100 Bangla sentences have been recorded from 5 (five) male speakers of different ages and 656 words have been presented in the 100 Bangla sentences. So, the speech database contains 500 Bangla sentences with 3280 words. All the algorithms and methods used in this research are implemented in MATLAB and the proposed system has been achieved the average segmentation accuracy of 90.58%.