Concise

Bengali_ASR and TIMIT dataset optimization and formation for ML training. Model generation with DNN and CNN in machine learning using Python which can be deployed in Flask, Azure and in industrial level software.

• Using SincNet technique Machine learning script generation and training the dataset.
• Different techniques like ARC Loss, Cos Loss implementation on dataset and find out the accuracy.
• Model generation using trained dataset and testing for accuracy.
• Final result plotting - example shown on the pictures above.
• Bengali_ARS global voice dataset manipulation (frequency and bitrate).
• Data formation according to the voice data tsv of Bengali_ASR.
• Sorting Data according to User ID and TIMIT Dataset .