OpenEARS/PocketSphinx Voice Recognition Optimization
I need a voice recognition expert to enhance a pretty simple app prototype's speech recognition system.
The prototype is an iOS Objective-C project, using OpenEARS/PocketSphinx as voice recognition library. We have three simple voice recognition situations, each having only a pretty small vocabulary with 4 words at max which have to be distinguished from each other. Each word has some variants so we have a complete vocabulary of about 20-30 words.
The recognition should especially work with the German language, so if you speak German it's a plus, but not a must.
The job consists of:
- Integration of a German language model into the system (there is a german language model which we can use)
- Creating an automated testing mode, getting hundreds of audio samples (recorded voice samples) as input and outputting the performance of the current model.
- Output of some statistic information which phonemes combinations have been detected mostly for each word.
- Tuning of the model to optimize the performance of the detection (at least 90%).
To apply you need to have an at least basic understanding of voice recognition and language models. Having experience with OpenEARS, RapidEARS or PocketSphinx is a great plus. You need to be able to write Objective-C code.
Please provide same samples of related work with your application.