We are using speech recognition in our VR application : we record samples form the microphone and then we send the samples into the cloud (GoogleSpeechReco) to obtain the result.
It works fine on the Rift -first version- but the same application, using Unity 2018.4.3f1, has bad recognition result on the Quest. Words are missing most of the time or/and wrong words are recognized (the result is too bad to be usable).
We have no voice detection (so no treshold) : the start and stop of recording are manual.
We pronounce the same sentence and it is not well recognized on the Quest, and it is ok on the Rift.
The only difference we have seen is the recording sample frequency, on the Rift it is 48KHz and on the Quest 16Khz.
We have no problem of recording authorization on Quest (we ask Android permission and manifest is ok).
We have also tried to save in a file what we record from the microphone. The result when we listen the file sounds ok : volume is lower than the Rift, but the result is bad even if we amplify the volume.
Has someone already experienced this kind of problem on the Quest?
If you have any idea of what can produce this difference?
All ideas are welcome...
Thanks for your help,