Interacting with machines by human voice is called speech recognition systems. These speech recognition systems are built separately for each of the human language. The challenge of the implementing these systems are the changing rules and regulations of the human language. So, it is very much necessary to build a separate system for specific language. Many of the Latin languages, Indian languages and others have their own speech recognition system. Sindhi an official language of Sindh province is still lacking this system as no speech recognition system have been found (to the best of our knowledge). For the development of any speech recognition system a basic element of database is necessary where the rules and regulations have been defined so that the speech recognition system can work upon. A comprehensive database has been created for the purpose of Sindhi speech recognition for the Sindhi language and for the purpose various parts of Sindh have been visited to collect speech samples spoken by various subjects around Sindh province so that the versatile database for the Sindhi SRS can be built. The subjects were requested to speak selected single word, double words, sentences and the continuous speech. An analysis of the comprehensive speech database is presented in this paper.

Key words: Speech Synthesis, recording environment, Acoustic Model, Datasets.