Abstract: Methods, systems, and apparatus, including computer programs encoded on computer storage media, for objection classification. One of the methods includes: obtaining a first set of images of objects that have a likelihood of being at a property that satisfies a likelihood threshold; generating, for each object, a binary classifier from a set of images of the respective object; determining, using at least one of the binary classifiers, that an image of an unknown object was classified as an object from the objects; in response to determining, using the binary classifiers, that the image of the unknown object was classified as an object from the objects, selecting a second set of images of unknown objects that does not include the image; and generating a multiclass classifier for use classifying objects using i) the first set as respective classes and ii) the second set that does not include the image.

Abstract: Methods, systems, and apparatus, including computer programs encoded on computer storage media, for fast user enrollment for facial recognition using face clustering. One of the methods includes identifying, from a set of face images of faces, clusters of face images, where the clusters of face images include a particular cluster; receiving, from a device, an indication that the particular cluster includes a first subcluster of face images that depict a first person and a second subcluster of face images that depict a second person; in response to receiving the indication, determining that a number of face images in the first subcluster of face images that depict the first person does not satisfy an enrollment criteria; identifying another cluster of face images that depict the first person; and enrolling, in a facial recognition database, the first person using the other cluster of face images.

Abstract: A method and apparatus for activity tracking based on time domain and frequency domain processing are disclosed. Embodiments according to the present invention are used to improve the accuracy of activity detection and step counting. The activity tracking starts from sample collection to generate 3-D accelerometer data. By pre-processing, the 3-D accelerometer data is calibrated and filtered. Then, the dominant component is calculated and statistical attributes or features used for activity detection are extracted. The statistical attributes are derived from time domain sensor data, frequency domain transformed data, or both. A classifier is developed using representative training data set. The activity detector determines the current activity status based on the statistical attributes and the classifier. To further refine the activity, post-processing is performed on the activity status.

Abstract: A system for counting steps comprising a 3-D accelerometer is disclosed. The system also includes a pre-processor module coupled to the 3-D accelerometer and a dominant component computation unit coupled to the pre-processor module. The dominant component computation unit is configured to identify a dominant component in an output of the 3-D accelerometer. The system further includes a step counter for counting a number of steps using the output of the dominant component computation unit. The step counter includes a Fast Fourier Transform (FFT) module and a direct current (DC) remover module to remove a static component from the output of the FFT module. The step counter also includes a derivative filter and a zero crossing detector.

Abstract: A system for counting steps comprising a 3-D accelerometer is disclosed. The system also includes a pre-processor module coupled to the 3-D accelerometer and a dominant component computation unit coupled to the pre-processor module. The dominant component computation unit is configured to identify a dominant component in an output of the 3-D accelerometer. The system further includes a step counter for counting a number of steps using the output of the dominant component computation unit. The step counter includes a Fast Fourier Transform (FFT) module and a direct current (DC) remover module to remove a static component from the output of the FFT module. The step counter also includes a derivative filter and a zero crossing detector.

Abstract: A method and apparatus for activity tracking based on time domain and frequency domain processing are disclosed. Embodiments according to the present invention are used to improve the accuracy of activity detection and step counting. The activity tracking starts from sample collection to generate 3-D accelerometer data. By pre-processing, the 3-D accelerometer data is calibrated and filtered. Then, the dominant component is calculated and statistical attributes or features used for activity detection are extracted. The statistical attributes are derived from time domain sensor data, frequency domain transformed data, or both. A classifier is developed using representative training data set. The activity detector determines the current activity status based on the statistical attributes and the classifier. To further refine the activity, post-processing is performed on the activity status.