estheroduntan/SPEECH_RECOGNITION_YORUBA

This is a project that deals with the transcription of a local language named Yoruba from the South West, Nigeria, Africa generated in .txt file and transcribed into .wav file for a Text To Speech Task # AMMI_2020_SPEECH_RECOGNITION_PROJECT This is a project that deals with the transcription of a local language named Yoruba from the South West, Nigeria, Africa generated in .txt file and transcribed into .wav file for a Text To Speech Task.Text to Speech Task was performed using customized audio data. A Case study of Yoruba Language; Speech Recognition is is an interdisciplinary subfield of computer science and computational linguistics that develops methodologies and technologies that enable the recognition and translation of spoken language into text by computers. There are three categories of these concepts namely: Text To Speech(TTS); Automatic Speech Recognition(ASR) and Speech To Text(STT). # DATA COLLECTION: The following stteps were taken in collecting data for the TTS 1. transcribing of text into an electronic format of .txt; the .txt file were organised into sessions, each session comprises of minimum of fifty(50) senetences. 2. Downloading and installing of the LIG Aikuma API on an Android phone, to enable the transcription of the .txt file to audio format of .wav. 3. On the LIG Aikuma API, the elicitation by Text was performed on each session of .txt file; which serves as input and the output were .json, .wav and linker file. 4. The output from the Android API was transferred to the laptop for preprocessing. 5. the audio files(.wav) were timed using vlc application and it reported a total of 2hrs 26 seconds. # TEXT PRE-PROCESSING Text preprocessing is an act of cleaning and preparing text data. In this project, the pre-processing was performed by developing a python script named: generate.py. This script was used to define the set of characters used in the yoruba language; it was also used to remove corrupted files in the data set, it was used to derive the all_session.txt file which was used for the model implementation. The all_session.txt file was tokenized and encoded into numeric format. The dataset was splitted into: train, validation and test …