Presentation is loading. Please wait.

Presentation is loading. Please wait.

Pocket, Bag, Hand, etc. - Automatically Detecting Phone Context through Discovery Emiliano Miluzzoy, Michela Papandreax, Nicholas D. Laney, Hong Luy, Andrew.

Similar presentations


Presentation on theme: "Pocket, Bag, Hand, etc. - Automatically Detecting Phone Context through Discovery Emiliano Miluzzoy, Michela Papandreax, Nicholas D. Laney, Hong Luy, Andrew."— Presentation transcript:

1 Pocket, Bag, Hand, etc. - Automatically Detecting Phone Context through Discovery Emiliano Miluzzoy, Michela Papandreax, Nicholas D. Laney, Hong Luy, Andrew T. Campbelly Presented by: Lulwah Alkwai

2 Introduction Discovery Framework Phone Sensing Context Design System Implementation Preliminary System Evaluation Conclusion

3 Introduction What is “phone sensing context”? The position of the phone carried by a person (e.g. in the pocket, hand, backpack, arm,...) in relation to the event being sensed. It is a fundamental building block for new distributed sensing application built on mobile phones. Observation has grown out of implementation of CenceMe and SoundSense.

4 CenceMe : CenceMe : Is a personal sensing system that enables members of social networks to share their sensing presence with their buddies in a secure manner. Is a personal sensing system that enables members of social networks to share their sensing presence with their buddies in a secure manner.

5 SoundSense : Top end mobile phones include a number of specialized (e.g., accelerometer, compass, GPS) and general purpose sensors (e.g., microphone, camera) that enable new people-centric sensing applications.

6 Discovery Framework Phone sensing Context Design System implementation

7 Phone Sensing Context AccurateRobust Low duty cycle

8 Design Using the entire suite of sensing modalities available on a mobile phone to provide enough data features for context discovery at low cost and for increased accuracy and robustness. Using the entire suite of sensing modalities available on a mobile phone to provide enough data features for context discovery at low cost and for increased accuracy and robustness.

9

10 System Implementation Feature selection: 1st-19 th : Audio signal classification problems 20 th : Power of audio signal/raw audio data 21 st,22 nd : Mean and standard deviation 23 rd : # of times exceeds a certain points TrainingPredictions

11 (a) FFT power of an audio clip,when the phone inside the pocket (b) FFT power of an audio clip,when the phone outside the pocket (c) Count the number of times the FFT power exceeds the threshold

12 Preliminary System Evaluation The result highlight that the audio modality is effective in detecting the in/out of pocket context with reasonable accuracy. The result highlight that the audio modality is effective in detecting the in/out of pocket context with reasonable accuracy. IN/OUT POCKET A: GMM B: SVM C: GMM TRAINING AND EVALUATING INDOOR D: SVM TRAINING AND EVALUATING OUTDOOR E: SVM TRAINING AND EVALUATING INDOOR F: SVM TRAINING OUTDOOR AND EVALUATING INDOOR G: GMM TRAINING USING ONLY MFCC H: SVM TRAINING USING ONLY MFCC

13 Conclusion Initial implementation looks promising, has potential, when implemented in its full form to become a core component of future mobile sensing systems. Initial implementation looks promising, has potential, when implemented in its full form to become a core component of future mobile sensing systems.

14 Thank you


Download ppt "Pocket, Bag, Hand, etc. - Automatically Detecting Phone Context through Discovery Emiliano Miluzzoy, Michela Papandreax, Nicholas D. Laney, Hong Luy, Andrew."

Similar presentations


Ads by Google