Speech recognition gives the text output at a given voice, in short it is speech to text (STT) conversion. It is useful for deaf, mute and disabled people. This project aims to improve the efficiency of speech recognition accuracy. Development of the voice recognition system with its own dictionary, in order to improve the efficiency of the voice recognition system. Errors generally vary not only in number, but also have different degrees of impact on the optimization of a set of acoustic models. It is important to correct errors in speech recognition results to increase the performance of a speech recognition system. Errors are detected and corrected based on the database learned from pairs of erroneous-correct statements. When the speech recognition system runs, it displays the values ​​and errors of the references and assumptions. By balancing errors, we can improve the accuracy of speech recognition. By removing silence from the speech signal, we can improve speech accuracy. Say no to plagiarism. Get a tailor-made essay on “Why Violent Video Games Should Not Be Banned”? Get an Original Essay Speech recognition is a process of converting spoken words into text. Speech recognition analyzes an acoustic speech signal to identify the linguistic message. Speech recognition systems compare spoken words and text and then give accuracy. These recognition systems play a vital role in facilitating daily activities. Speech recognition applications include voice dialing, call routing and content-based spoken audio search, data entry, structured document preparation, speech processing in text and in aircraft cockpits. In addition to this, the voice recognition system can be used for people with visual impairments or paralyzed hands. In underdeveloped countries where literacy rates are low, this can provide a mechanism for accessing information to people unable to read and write as well as those who may be literate but not computer skilled . Speech recognition is defined as the ability of a computer to understand spoken commands or responses and is an important factor in human-computer interaction. SR has been available for many years, but has not been practical due to the high cost of applications and computing resources. SR has seen significant growth in telephony and voice-to-text applications. Increasing the efficiency of workers who perform intensive typing, assist people with disabilities, and manage call centers by reducing staffing costs, shows the benefits of voice recognition. Speech recognition is the process by which a computer identifies spoken words. Basically, this means talking to your computer and making it correctly recognize what you say. It is simply a signal to symbol transformation, i.e. it takes speech as input and gives text as output. Recognition models: Speaker-dependent: Speech recognition systems that can only recognize users' speech that they are trained to understand are called speaker-dependent speech recognition devices. Limited to the understanding of selected speakers. Speaker-independent: Speech recognition software that recognizes a variety of speakers, without any training, is called speaker-independent speech recognition system. Hidden Markov model:Each speech recognition system is associated with the Hidden Markov Model: A Hidden Markov Model is a probabilistic state machine that can be used to model and recognize speech. Consider the speech signal as a sequence of observable events generated by the mechanical speech production system that transitions from one state to another during speech production. The term hidden refers to the fact that the state of the system (i.e., the configuration of the speech articulators) is not known to the observer of the speech signal. Speech recognition systems use HMMs to model each sound unit of language. In an HMM, each state is associated with a probability distribution that measures the probability of events generated by the state. These distributions are known as output or observation probability distributions. Each state is also associated with a set of transition probabilities. Given the current state, transition probabilities model the probability that the system will be in a certain state when the next observation is produced. Generally, Gaussian distributions are used to model the output distribution of each HMM state. Transition probabilities determine how quickly the model transitions from one state to another, giving the model some flexibility with respect to sound units that can vary in duration. HMM = (?, A, B) ? = the vector of initial state probabilities A = the state transition matrix B = the confusion matrix The definitions of HMMs, there are three problems of interest: The evaluation problem: The forward-backward algorithm is used to find the probability that the model generated the observations for a given model and sequence of observations. The decoding problem: The Viterbi algorithm can find the most likely state sequence in the model that produced the observation for a given model and the sequence of observations. The learning problem: The Baum-Welch algorithm finds the model parameters so as to obtain the maximum probability of generating the observations for a given model and a sequence of observations. (A) Direct algorithm: The direct algorithm calculates all possible length state sequences that generate an observation sequence, then sums all probabilities. The probability of each path is the product of the state sequence probability and the joint probability along the path. (B) Viterbi algorithm: the direct algorithm calculates the probability that an HMM generates an observation sequence by summarizing the probabilities of all possible paths, so it does not provide the best path or the best sequence of states. In many applications it is desirable to find such a path. Finding the best is the cornerstone of the pursuit of continuous speech recognition. Since the state sequence is hidden in the HMM framework, the most widely used criterion is to find the state sequence that has the highest probability of being taken when generating the observation sequence. The Viterbi algorithm can be considered as dynamic programming applied to HMM. or as a modified transmission algorithm. Instead of summarizing the probabilities of different paths arriving at the same destination state, the Viterbi algorithm selects and remembers the best path. (C) Baum-Welch Algorithm: It is also known as forward-backward algorithm used to model the observations in the training data via HMM parameters. This algorithm is a kind of EM (Expectation Maximization) algorithm that goes through the data first in a forward pass and then in a backward pass. During each pass, we adjust a set of probabilities to maximize the.