Abstract: Sleep conditions such as moderate-to-severe sleep apnea can be assessed using a multi-night assessments. A respiration signal (e.g., acquired from a sensor strip) can be processed via a computing device. The respiration signal can be segmented and the segments can be classified to identify one or more apnea/hypopnea events. In some examples, some of the segments can be normalized such that each segment input for classification can be of the same size. The identified one or more apnea/hypopnea events can be used to estimate a nightly parameter indicative of a severity of (or presence of) sleep apnea. The nightly parameters from a multi-night period can be used to estimate a multi-night parameter indicative of the severity of (or presence of) sleep apnea. In some examples, quality checks can be performed to filter out some data (e.g., to exclude data from entire nights or exclude a portion of data from individual nights).

Abstract: A system for selecting different aspects of data objects to be matched with similar aspects of other data objects. A user inputs a search data object and a value. A neural network computes features for the search object at multiple layers that correspond to different aspects of the object. A descriptor is generated for the search object from features output at a layer position of the neural network determined from the value. The descriptor is compared to corresponding descriptors for objects in a collection to select objects that include aspects similar to an aspect of the search object. The user can change the value to view different objects that include aspects similar to other aspects of the search object. Thus, the user can explore different aspects of an object to find objects that include aspects similar to the aspect of the object that the user is interested in.

Abstract: Modifying operation of an intelligent agent in response to facial expressions and/or emotions.

Abstract: Modifying operation of an intelligent agent in response to facial expressions and/or emotions.

Abstract: In some examples, an individually-pruned neural network can estimate blood pressure from a seismocardiogram (SMG). In some examples, a baseline model can be constructed by training the model with SMG data and blood pressure measurement from a plurality of subjects. One or more filters (e.g., the filters in the top layer of the network) can be ranked by separability, which can be used to prune the model for each unseen user that uses the model thereafter, for example. In some examples, individuals can use individually-pruned models to calculate blood pressure using SMG data without corresponding blood pressure measurements.

Abstract: Modifying operation of an intelligent agent in response to facial expressions and/or emotions.

Abstract: Modifying operation of an intelligent agent in response to facial expressions and/or emotions.

Abstract: A “Food Logger” provides various approaches for learning or training one or more image-based models (referred to herein as “meal models”) of nutritional content of meals. This training is based on one or more datasets of images of meals in combination with “meal features” that describe various parameters of the meal. Examples of meal features include, but are not limited to, food type, meal contents, portion size, nutritional content (e.g., calories, vitamins, minerals, carbohydrates, protein, salt, etc.), food source (e.g., specific restaurants or restaurant chains, grocery stores, particular pre-packaged foods, school meals, meals prepared at home, etc.). Given the trained models, the Food Logger automatically provides estimates of nutritional information based on automated recognition of new images of meals provided by (or for) the user. This nutritional information is then used to enable a wide range of user-centric interactions relating to food consumed by individual users.

Abstract: A “Food Logger” provides various approaches for learning or training one or more image-based models (referred to herein as “meal models”) of nutritional content of meals. This training is based on one or more datasets of images of meals in combination with “meal features” that describe various parameters of the meal. Examples of meal features include, but are not limited to, food type, meal contents, portion size, nutritional content (e.g., calories, vitamins, minerals, carbohydrates, protein, salt, etc.), food source (e.g., specific restaurants or restaurant chains, grocery stores, particular pre-packaged foods, school meals, meals prepared at home, etc.). Given the trained models, the Food Logger automatically provides estimates of nutritional information based on automated recognition of new images of meals provided by (or for) the user. This nutritional information is then used to enable a wide range of user-centric interactions relating to food consumed by individual users.

Abstract: A “Food Logger” provides various approaches for learning or training one or more image-based models (referred to herein as “meal models”) of nutritional content of meals. This training is based on one or more datasets of images of meals in combination with “meal features” that describe various parameters of the meal. Examples of meal features include, but are not limited to, food type, meal contents, portion size, nutritional content (e.g., calories, vitamins, minerals, carbohydrates, protein, salt, etc.), food source (e.g., specific restaurants or restaurant chains, grocery stores, particular pre-packaged foods, school meals, meals prepared at home, etc.). Given the trained models, the Food Logger automatically provides estimates of nutritional information based on automated recognition of new images of meals provided by (or for) the user. This nutritional information is then used to enable a wide range of user-centric interactions relating to food consumed by individual users.

Abstract: A system and method to determine pulse transit time using a handheld device. The method includes generating an electrocardiogram (EKG) for a user of the handheld device. Two portions of the user's body are in contact with two contact points of the handheld device. The method also includes de-noising the EKG to identify a start time when a blood pulse leaves a heart of the user. The method further includes de-noising a plurality of video images of the user to identify a pressure wave indicating an arterial site and a time when the pressure wave appears. Additionally, the method includes determining the PTT based on the de-noised EKG and the de-noised video images.

Abstract: Aspects of the subject disclosure are directed towards a video-based pulse/heart rate system that may use motion data to reduce or eliminate the effects of motion on pulse detection. Signal quality may be computed from (e.g., transformed) video signal data, such as by providing video signal feature data to a trained classifier that provides a measure of the quality of pulse information in each signal. Based upon the signal quality data, corresponding waveforms may be processed to select one for extracting pulse information therefrom. Heart rate data may be computed from the extracted pulse information, which may be smoothed into a heart rate value for a time window based upon confidence and/or prior heart rate data.

Abstract: A “Food Logger” provides various approaches for learning or training one or more image-based models (referred to herein as “meal models”) of nutritional content of meals. This training is based on one or more datasets of images of meals in combination with “meal features” that describe various parameters of the meal. Examples of meal features include, but are not limited to, food type, meal contents, portion size, nutritional content (e.g., calories, vitamins, minerals, carbohydrates, protein, salt, etc.), food source (e.g., specific restaurants or restaurant chains, grocery stores, particular pre-packaged foods, school meals, meals prepared at home, etc.). Given the trained models, the Food Logger automatically provides estimates of nutritional information based on automated recognition of new images of meals provided by (or for) the user. This nutritional information is then used to enable a wide range of user-centric interactions relating to food consumed by individual users.

Abstract: In exemplary implementations, this invention comprises apparatus for retinal self-imaging. Visual stimuli help the user self-align his eye with a camera. Bi-ocular coupling induces the test eye to rotate into different positions. As the test eye rotates, a video is captured of different areas of the retina. Computational photography methods process this video into a mosaiced image of a large area of the retina. An LED is pressed against the skin near the eye, to provide indirect, diffuse illumination of the retina. The camera has a wide field of view, and can image part of the retina even when the eye is off-axis (when the eye's pupillary axis and camera's optical axis are not aligned). Alternately, the retina is illuminated directly through the pupil, and different parts of a large lens are used to image different parts of the retina. Alternately, a plenoptic camera is used for retinal imaging.

Abstract: A system and method to determine pulse transit time using a handheld device. The method includes generating an electrocardiogram (EKG) for a user of the handheld device. Two portions of the user's body are in contact with two contact points of the handheld device. The method also includes de-noising the EKG to identify a start time when a blood pulse leaves a heart of the user. The method further includes de-noising a plurality of video images of the user to identify a pressure wave indicating an arterial site and a time when the pressure wave appears. Additionally, the method includes determining the PTT based on the de-noised EKG and the de-noised video images.

Abstract: In exemplary implementations, this invention comprises apparatus for retinal self-imaging. Visual stimuli help the user self-align his eye with a camera. Bi-ocular coupling induces the test eye to rotate into different positions. As the test eye rotates, a video is captured of different areas of the retina. Computational photography methods process this video into a mosaiced image of a large area of the retina. An LED is pressed against the skin near the eye, to provide indirect, diffuse illumination of the retina. The camera has a wide field of view, and can image part of the retina even when the eye is off-axis (when the eye's pupillary axis and camera's optical axis are not aligned). Alternately, the retina is illuminated directly through the pupil, and different parts of a large lens are used to image different parts of the retina. Alternately, a plenoptic camera is used for retinal imaging.