Spoken Words (Voice) Recognition TensorFlow / Open Source

Project only for those who have many experience in Voice Recognition / Tensorflow / Recognition Libraries. We seek for a ML engineer who has lots of experience with TensorFlow or other Open Source 'recognition' libraries. We need as soon as possible a 99% accurate Speech / Voice Recognition that has the ability to be implemented on Android and iOS and running on the background while the app is working on the background - not in use - without a button being pressed. You may not use the Android / iOS built-in libraries, we need our own Library. The training must be done as well. Word training concept: 1. We provide you certain - +/- 200 words. Lots of training is required, which can be partially automatically. When these words are spoken by the user, a trigger must be transmitted. 2. App users may speak own words. An app user must speak one word about ten times, to help the training. Once recognized. When these words are spoken by the user, a trigger must be transmitted. 3. Pinch / Frequency when the frequency is higher (loudness is higher) a trigger must be transmitted as well, in combination with the words - we increase the alert accuracy. You are allowed to use Open Source / Partially or Fully Trained models, as long as we have full access to the eventual source code. Budget for phase one: $1000 We need it very fast! - So contact us fast. Other phases have a larger budget - based on the results of Phase One. // Second phase also a SDK which should operate on Linux based Arm devices that have a microphone in the USB. But that is part two. // Third phase includes many more words, in many different languages. // Fourth phase includes more features.