Speech Recognition

The design of speech recognition is a developing technology, but the applications are currently being used by companies to answer and direct calls without the presence of an operator and human resources. Conceptually the idea of voice recognition has many applications in the military and medical industries. Accuracy is difficult with factors such as dialect and the environment of the speaker. In 1954 Manfred Schroder joined Bell Labs to research speech and computer programming and came up with ways in which audio can be transformed through certain formats. He made over forty relative patents. Twenty two years later research and analysis of mel-frequency cepstral coefficients enables further development of speech recognition from mathematical concepts. Defining and experimentation of the cepstrum lays a foundation for programs but is not robust (not very accurate unless in specific conditions). In 1993 Lawrence Rabiner publishes “Fundamentals of Speech Recognition” and in 1995 intense publications on robustness are written by Jean-Claude Junqua. Jean-Claude writes about studies done by the IEEE Signal Processing Society and the future of PC enabling voice command. Problems and goals are assessed bianually at the IEEE workshop on Automatic Speech Recognition and Understanding. The 10th annual meeting was held December 9th through 13th last year. Presently, Dr. Joseph Olive heads GALE project for DARPA. GALE is largest ongoing project in speech recognition and related areas. The ultimate goal of Global Autonomous Language Exploitation is not just having voice to text computer software but also to store large amounts after any given time of speech. The developing programs will be able to analyze and translate the audio accurately into “consolidated information in easy-to-understand forms.”