I'm a Ph.D. - Pediatric , BUT i do have some programming knowledge
mysql/php and some linux Commands :).
i'm looking to re recognize "baby crying sound" in order to know the baby
needs. We need it to deal with the sound as an event, not as music.
I did record a lot of baby crying sounds and identify each sample with
the "baby status" , is he / she hungry , tired..etc. As learning sample
pattern. In order to let the system learn from the samples
Can i make Sphinx learn from my sample in order to create a unique
language.
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
Sure, you can use CMUSphinx to recognize different sounds in baby crying. You need to prepare the database with proper markup and train the acoustic model on your database. There are also issues with creation of the transcription for the markup. you need to understand which units in HMM do you want to have. Whole cry with a multiple repeating patterns must be a sequence of HMM units, not a single unit.
Also, it's not clear if single GMM for different babies will work, i.e. if it will be possible to interpolate units of different babies into single HMM detector. In case you want a baby-independent cry recognition.
Feature extraction parameters must be changed too in order to extract features from cry.
Overall, in order to apply CMUSphinx or any other HMM toolkit you need to understand HMM model and how is it used in recognition. A Rabiner's tutorial on HMM should get you started. It doesn't really matter which HMM framework to use, it more matters how do you apply HMM framework for your task.
If you can provide cry examples, I can probably be more specific on what you need to do.
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
I'm a Ph.D. - Pediatric , BUT i do have some programming knowledge
mysql/php and some linux Commands :).
i'm looking to re recognize "baby crying sound" in order to know the baby
needs. We need it to deal with the sound as an event, not as music.
I did record a lot of baby crying sounds and identify each sample with
the "baby status" , is he / she hungry , tired..etc. As learning sample
pattern. In order to let the system learn from the samples
Can i make Sphinx learn from my sample in order to create a unique
language.
Hello beshoo
Sure, you can use CMUSphinx to recognize different sounds in baby crying. You need to prepare the database with proper markup and train the acoustic model on your database. There are also issues with creation of the transcription for the markup. you need to understand which units in HMM do you want to have. Whole cry with a multiple repeating patterns must be a sequence of HMM units, not a single unit.
Also, it's not clear if single GMM for different babies will work, i.e. if it will be possible to interpolate units of different babies into single HMM detector. In case you want a baby-independent cry recognition.
Feature extraction parameters must be changed too in order to extract features from cry.
Overall, in order to apply CMUSphinx or any other HMM toolkit you need to understand HMM model and how is it used in recognition. A Rabiner's tutorial on HMM should get you started. It doesn't really matter which HMM framework to use, it more matters how do you apply HMM framework for your task.
If you can provide cry examples, I can probably be more specific on what you need to do.