Phonetic Speech Analysis for Speech to Text Conversion

This paper presents a description of the work done on phonetic speech analysis. The work aims in generating phonetic codes of the uttered speech in training-less, human independent manner. This work is guided by the working of ear in response to audio signals. The Devnagri script inspires the work presented. The Devnagari script classifies and arranges 46 phonemes in a scientific manner based on the process of its generation. The work at present focuses on identifying the class (varna) of the phoneme as specified by the Devnagari script. More work is needed to identify the variant of the class identified. Phoneme code thus generated can be used in an application specific way. This work also explains and proves the scientific arrangement of the Devnagari script. This work tries to segment speech into phonemes and identify the phoneme using simple operations like differentiation, zero-crossing calculation and FFT.