1

i would like to create an emotion speech recognition application in android which will detect the emotion of the users through getting the sound of their voice. is it possible to make this kind of app in android?if yes, can you give me links that will help me to make this possible. thanks!

Hazel
  • 53
  • 1
  • 8
  • Google has a speech recognition system. I don't know if Google has emotions though. – Jeremy D Jul 25 '12 at 13:44
  • @JeremyD, I meant Google Search by the way =) – Egor Jul 25 '12 at 13:45
  • 1
    sounds like homework to me... – 10s Jul 25 '12 at 13:48
  • yes i did. actually right now i tried using the recognizerintent which now getting the speech of the user but im now stacked and i don't know what method i should used next and how am i going to implement the feature extraction. do you think it's possible? – Hazel Jul 25 '12 at 13:49
  • Hard homework then..:) On what are you stuck exactly? What results do you have? Any example? Data? – Jeremy D Jul 25 '12 at 13:49
  • @JeremyD haha yes Google supports speech recognition and i've already implemented it but my problem now is i don't know wat are the things i need to do after implementing the Google speech recognition. i mean like with the feature extraction and other algorithms possible to use,classifications. do u know any links which will help me? – Hazel Jul 25 '12 at 13:57
  • If you just Google for your requirement, it will give you many open source API's implemented in Java(and some specifically for Android). For your specific requirement(emotion recognition) you might have to explore some of these API's and work on them. Your idea sounds innovative. Emotions in a voice can be categorized using pitch of sound, intervals between pauses and the words being used. – Paramvir Singh Jul 25 '12 at 13:58
  • As suggested in some answers, there are many academic papers exploring this. I thought this might be helpful too, it is more of a business-centric article summarizing some current thinking about emotion detection in voice applications - http://www.callcentrehelper.com/does-emotion-detection-really-work-9047.htm – Michael Levy Jul 25 '12 at 14:03
  • this is indeed an interesting topic. – Jeremy D Jul 25 '12 at 17:30
  • See also http://stackoverflow.com/questions/7673736/emotion-detection-in-speech – Nikolay Shmyrev Nov 09 '13 at 10:55

2 Answers2

4

In any software application, you need three things:

Inputs --> Processing --> Outputs

So you will have to split this project into these three categories:

  1. INPUTS: Recording voice data
  2. PROCESSING: Run an algorithm on the recorded voice file for emotion detection
  3. OUTPUTS: Write an emotion to the screen

To achieve number one is quite simple: http://www.benmccann.com/blog/android-audio-recording-tutorial/

The second part is more difficult, you will have to do some research and read some papers: http://www.google.co.za/url?sa=t&rct=j&q=&esrc=s&source=web&cd=1&ved=0CGoQFjAA&url=http%3A%2F%2Fcs229.stanford.edu%2Fproj2011%2FVeitZilberMordkovich-DetectingEmotionInHumanSpeech.pdf&ei=uvgPUMXRIcW3hQfA64DgAw&usg=AFQjCNEwLeC9R8jL-HOoBS6QWu-6dbRabQ

http://www.google.co.za/url?sa=t&rct=j&q=&esrc=s&source=web&cd=2&ved=0CGwQFjAB&url=http%3A%2F%2Fcs229.stanford.edu%2Fproj2007%2FShahHewlett%2520-%2520Emotion%2520Detection%2520from%2520Speech.pdf&ei=uvgPUMXRIcW3hQfA64DgAw&usg=AFQjCNFGwlKlcNyjG4uK4Hbmyy6emuYpIA

The third part is the easiest - and will be left to you as an exercise.

Ben McCann
  • 17,641
  • 23
  • 79
  • 99
Quintin Balsdon
  • 5,023
  • 9
  • 51
  • 93
  • thank you for this links i'll try the first one with the recording voice data. actually im now using the recognizerintent for the speech recognition but i'll try this link you've suggested. thanks again – Hazel Jul 25 '12 at 14:14
  • and oh yes i have also gone reading and researching more on about emotions and emotion detection in speech recognition. and i am now trying to implement them through creating them in code – Hazel Jul 25 '12 at 14:21
3

you could also do a very simple emotion detection on the words returned by Google's speech recognition engine. this does not capture the emphasis with which words are spoken but is currently probably the much more realistic approach.

Anno2001
  • 1,333
  • 12
  • 17