الفهرس | Only 14 pages are availabe for public view |
Abstract Automatic speech recognition (ASR) is recognized as the independent computer-driven transcription which changes talked language into legible text. ASR lets a computer to get the words from a person that speaks into a microphone and alter them into written text. ASR is used in a lot of applications in real-time. The phonetic features and articulation of various sounds are necessary for the classification into separate categories. This classification of sounds can be applied for applications like speaking rate evaluation, speech recognition, phone recognition, and language recognition. In this thesis, there are four hybrid techniques based on the acoustic-phonetic approach and pattern recognition approach that are used to emphasize the principle idea of this research. The first hybrid model is constructed of Static State, structured Hidden Markov Model, Gaussian Mixture, Mel scaled Best Tree Image, Convolution Neural network, Vector Quantization (SS-HMM-GM-MBTI-CNN-VQ). The second hybrid model is constructed of Dynamic State, structured Hidden Markov Model, Gaussian Mixture, Mel scaled Best Tree Image, Convolution Neural network, Vector Quantization (DS-HMM-GM-MBTI-CNN-VQ). The third hybrid model is constructed of Static State, structured Hidden Markov Model, Gaussian Mixture, Mel scaled Best Tree Image, Convolution Neural network (SS-HMM-GM-MBTI-CNN). The fourth hybrid model is constructed of Dynamic State, structured Hidden Markov Model, |