Sound Detection System

Demo 1: Differentiating scream and non-scream sound

Demo 2: Automatically dial a specified phone number

Demo 3: Showing the detection decisions on notification page

  Android installation (.apk) file 1: With phone, Email and SMS capability
Android installation (.apk) file 2: Without phone, Email and SMS capability
Eclipse project and source code

---

Funding Source: Motorola Solutions Foundation

---

Motivation of Work

Surveillance and monitoring for security purposes have become increasingly important in today's society. Traditionally, surveillance is achieved by recording the scenes via video cameras. However, taking video is inappropriate (e.g., in a washroom) or even impossible (e.g., insufficient lighting) under some situations. Audio is a viable alternative under such situations. In fact, audio is more effective than video in detecting the events that posses unique acoustic signatures, e.g., screaming, crying, gunshots, door slam, etc. Moreover, video-based surveillance typically requires using fixed cameras to record the scenes, restricting the surveillance system to be localized in one place. However, audio-based surveillance can make effective use of mobile devices, allowing the surveillance system to be moved from one place to another easily. The recent ubiquity of smartphones has opened up new applications of sound event detection. For example, any smartphones can be turned into a personal security device, allowing users to detect any hazardous situations around them 24 hours a day. Abnormal sound events such as screaming can be detected and emergency phone calls can be automatically made. This kind of personal security device is particularly useful for children and women who are more vulnerable to assaults. The ability to detect scream sounds and body fall is also useful for the elderly.

Publications

1. Baiying Lei and M.W. Mak, "Robust scream sound detection via sound event partitioning", Multimedia Tools and Applications, March 2015, pp. 1-19, DOI: 10.1007/s11042-015-2555-z.
   
   
2. B.Y. Lei and M.W. Mak, "Sound-Event Partitioning and Feature Normalization for Robust Sound-Event Detection", International Conference on Digital Signal Processing, 2014, Paper Presentation File
   
   
3. M.W. Mak and S.Y. Kung, "Low-Power SVM Classifiers for Sound Event Classification on Mobile Devices", ICASSP'12, Kyoto, 2012, pp. 1985-1988. Paper

Example of Scream Sounds

---

M.W. Mak's Homepage

http://www.eie.polyu.edu.hk/~mwmak/mypage.htm