These days, voice to text recognition system based applications are growing progressively to deal with the daily life task. Many studies have been devised their voice to text systems in different languages. However, voice to text recognition system in the Sindhi language has not been devised yet. With this motivation, in this paper, we propose Sindhi voice to text recognition system based on Hidden Markov Model (HMM). We devise a novel Sindhi Recognition System based on Hidden Markov Model (SRS-HMM) and create a novel Sindhi dictionary to deal with users and their Quality of Service (QoS) requirements. The study formulates the combinatorial problem where minimum error-rate considered as a convex function and maximum accuracy formulated as a concave function. The objective of the study is to improve users input with different accents and enhance their accuracy and minimize the error rate. Simulation results show that, proposed SRS-HMM system gained near-optimal, e.g., 60% of different users with different accents during the performance.

   Keywords— Hidden Markov Model SRS-HMM Accent Training QoS.