I am currently testing the KWS mode/JSGF mode/ N-gram mode using various accoustic models (continuous, semi-continuous and ptm) on PI 2,3,4 as my master's thesis. My mission is to find the best way to acheive decent speech performance in terms of RTF, WER and CPU usage. Do you have any suggestions on how to set the optimum tresholds for every keyword for the kws mode? I have devised a formula for that and so far it has been performing well against the random treshold values. I need to collect more methods/ formulas or evidences in order to prove my hypothesis. Any suggestions as to how can I compare my hypothesis?
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
Hi Ismail. Pocketsphinx definitely works with RPI3. All you need to do is reduce the size of the vocabulary. You can try using the grammar mode for better performance
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
Hello Developers,
I am currently testing the KWS mode/JSGF mode/ N-gram mode using various accoustic models (continuous, semi-continuous and ptm) on PI 2,3,4 as my master's thesis. My mission is to find the best way to acheive decent speech performance in terms of RTF, WER and CPU usage. Do you have any suggestions on how to set the optimum tresholds for every keyword for the kws mode? I have devised a formula for that and so far it has been performing well against the random treshold values. I need to collect more methods/ formulas or evidences in order to prove my hypothesis. Any suggestions as to how can I compare my hypothesis?
Also, while testing the n-gram mode what parameters should I experiment with to acheive the accuracy? prameters such as -beam -wbeam -wip
Hello Aanand, sorry i want to ask a few things not about your questions.
Is it work using pocketsphinx to decode word in raspy 4 ? Because last time i use rapy 3b+ it gives me error because cpu pi 3b+ too slow
Thanks for your time, and i hope you success on your master's thesis
Hi Ismail. Pocketsphinx definitely works with RPI3. All you need to do is reduce the size of the vocabulary. You can try using the grammar mode for better performance
Last edit: ismail nur adli 2020-05-02
Last edit: ismail nur adli 2020-05-02
Sry accidentaly post 3 times. My bad
Last edit: ismail nur adli 2020-05-02
Hello People. I need your inputs. Any critisism or comments are welcome :)