Thus we have created a dataset by pre-recording sounds from 13 different sound sources and further generated morse code audio files from them. To the best of our knowledge there is no publicly available dataset that suits the aim of our problem. For training, we trained a C-RNN model with 2 Bi-directional LSTMs and CTC loss. In dataset curation, we pre-recorded audio files of different mediums and along with a text corpus, created a labeled morse code dataset. Our approach is broken up into two main phases: dataset curation and training. Output: english text corresponding to the inputted morse Approach More specifically, to develop such a morse code recognition system, we will decipher morse code audio into english text, irrespective of the sound source used to generate the morse code the system will be able to decipher the morse code into english text regardless of the sound source used to generate the input. This project's goal is to develop a system which will replicate this "human-like" behavior in deciphering morse code, i.e. Humans can decipher morse code in any given sound source, if they are first trained on a single source. In audio form, morse code can be generated via various methods, like tapping (on different materials such as bells, musical instruments, car honks, etc.). It is transmitted by on-off keying of an information carrying medium such as electric current, radio waves, visible light or sound waves. Morse code is a series of dots and dashes representing different text characters, digits and punctuation marks. ![]() ![]() Live Demo | Source Code | Project Video Introduction Tanmay Ghai, Ankur Garg, Revathi Mukkamala, Onur Orhan, Supriya Devalla Deciphering Morse Code Audio from Various Sound SourcesĬSCI 566: Deep Learning and its Applications, Fall 2020
0 Comments
Leave a Reply. |
AuthorWrite something about yourself. No need to be fancy, just an overview. ArchivesCategories |