Speech Recognition to Build Context: A Survey

In era Computer evolution many problems can be solved using computer vision and signal processing. These domains are typically Digitized in binary files like Images, Audio, and Videos. The translation, recognition and synthesis are required while understating the meaning of the binary content. The recognition process is also having many problems in case of audio processing. The missing context is the major reason in pattern-based matching. This is due to unclear or low-quality input, as well as training model on different frequencies but by using context some of the accuracy may improve. Context finding from binary files is a challenge as it works in temporal and space domain. Binary data like images contain special information, while audio files contain temporal information. Video files have both time and space domains. Updating context in the temporal domain, to find proper context from the audio corpus, speech recognition is applied. Over the time period, there are different models adapted like Hidden Markov Model (HMM), Rule Based models with fuzzy support, pattern-based models including machine learning techniques K-nearest neighbor, Support Vector Machine, also latest techniques like Artificial Neural Network (ANN). These technologies are typically included in Automatic Speech Recognition (ASR). ASR uses Language resources with any one of the above models. Here, an in-depth survey on ASR and available APIs. Technologies used to build APIs also discussed.