Resources: Sample data: (1) | Related publications: [under review]
Summary:
NRFAR is an algorithm designed to recognize foraging behavior of cattle in free-grazing environments. It analyzes a sound recording to recognize jaw-movement events and foraging activities. First, the algorithm performs the detection and classification of jaw-movement events (chew occurring in rumination (r), chew occurring in grazing (c), bites (b), and composed chew-bites (x)). Then, statistical features of previously recognized jaw movements are computed to determine the foraging activities (grazing and rumination bouts).
Usage description:
This web-demo requires an audio recording file as input (download examples). The input sound must be a WAV format audio file (16-bit, 44.1 kHz) between 2 and 15 min of duration. The WAV file could be one channel (mono) or two channel (stereo). The web-demo automatically converts the audio from stereo to mono.
If the value of the gain applied during the recording of the audio file is available, it could be used to improve the results.
Steps
1) Click “Choose file” and browse for the audio file to upload.
2) Click Upload to load the audio file in the web-demo. If either no file is provided nor Upload is clicked, an example audio file will be loaded by default.
3) Set the numerical value of the gain, if it is available. By default, it will be 1.
4) Click Submit to run the algorithm.
5) The output are two TXT format text files with timestamps and labels corresponding to:
– foraging activities (rumination and grazing bouts). Other labels will be assigned when no activity is identified.
– Timestamps and labels corresponding to jaw-movement events (r, c, b or x).
Both input (sound signal) and output (timestamps and labels) files can be imported in Audacity to get a graphical view of the recognition.