First of all, I am very new to sound recognition. Just started my searches on the subject several hours ago.
I am a freelance software developer and I am asked if I can develop an application on Raspberry Pi 3 that application recognize a crying baby and do some operations. I am able to do "do some operations" but I am trying to understand if I can do the main job.
Application shall need to know if baby is crying. It is not important why (hunger, pain, etc.) he/she's crying. I just need to be able to recognize that he/she is crying.
I have following questions with my very limited knowledge.
1- Can I use sphinx under Raspberry Pi 3 device?
2- Can I use sphinx with Lazarus (FreePascal) programming language?
3- Is there any trained library that I might use for my purpose?
4- I read a bit of http://cmusphinx.sourceforge.net/wiki/tutorialam. I have opportunity to collect crying baby sounds for training. Can I do a good training with less hours and less different babies than indicated in above tutorial? I really do not want to go in that direction if I can avoid it though.
I appreciate any help on that.
Thank & regards,
Ertan Küçükoğlu
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
cmusphinx is not supposed to be useful for this task, it is mainly designed for speech recognition, not for sound detection, you'd better look on some other software like Shazam or something.
You have to collect a lot of training data in any case.
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
I will check Shazam technology as much as I can. I know it as a closed source application. It may not be easy to find information about;
- How to do spectogram from baby cries?
- What is the algorithm I should be using for searching pattern in the trained data?
- How to do search on a trained database continuously? (Maybe, start when pitch Hz arrives at 200 or 250 which is considered as a baby cry beginning level?)
I appreciate any help on above questions.
Lastly, what mount of data I should be collecting for training? I have kind of an unlimited source. I just need to know approximate amount so that I can start that job as soon as possible. Please consider that application will be country specific if that helps to reduce training data.
Thanks & regards.
Last edit: Ertan Küçükoglu 2017-04-23
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
Hello Ertan,
I'm also interested in baby cry analysis.
I'll begin my work in September or October.
Have you some progress in your work about baby cries?
Can you share your baby cries?
Thanks , Regards, Michel
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
Hello,
First of all, I am very new to sound recognition. Just started my searches on the subject several hours ago.
I am a freelance software developer and I am asked if I can develop an application on Raspberry Pi 3 that application recognize a crying baby and do some operations. I am able to do "do some operations" but I am trying to understand if I can do the main job.
Application shall need to know if baby is crying. It is not important why (hunger, pain, etc.) he/she's crying. I just need to be able to recognize that he/she is crying.
I have following questions with my very limited knowledge.
1- Can I use sphinx under Raspberry Pi 3 device?
2- Can I use sphinx with Lazarus (FreePascal) programming language?
3- Is there any trained library that I might use for my purpose?
4- I read a bit of http://cmusphinx.sourceforge.net/wiki/tutorialam. I have opportunity to collect crying baby sounds for training. Can I do a good training with less hours and less different babies than indicated in above tutorial? I really do not want to go in that direction if I can avoid it though.
I appreciate any help on that.
Thank & regards,
Ertan Küçükoğlu
cmusphinx is not supposed to be useful for this task, it is mainly designed for speech recognition, not for sound detection, you'd better look on some other software like Shazam or something.
You have to collect a lot of training data in any case.
Hello Nikolay,
Thank you for your feedback.
I will check Shazam technology as much as I can. I know it as a closed source application. It may not be easy to find information about;
- How to do spectogram from baby cries?
- What is the algorithm I should be using for searching pattern in the trained data?
- How to do search on a trained database continuously? (Maybe, start when pitch Hz arrives at 200 or 250 which is considered as a baby cry beginning level?)
I appreciate any help on above questions.
Lastly, what mount of data I should be collecting for training? I have kind of an unlimited source. I just need to know approximate amount so that I can start that job as soon as possible. Please consider that application will be country specific if that helps to reduce training data.
Thanks & regards.
Last edit: Ertan Küçükoglu 2017-04-23
Hello Ertan,
I'm also interested in baby cry analysis.
I'll begin my work in September or October.
Have you some progress in your work about baby cries?
Can you share your baby cries?
Thanks , Regards, Michel