Abstract: A system can map occupancy of objects. The system may generate an occupancy representation of an object based on a point cloud of the object. The system may generate first geometric entities based on the point cloud. Each first geometric entity contains one or more points in the point cloud. The system may also generate one or more second geometric entities, each of which contains one or more first geometric entities. The occupancy representation of the object includes the one or more second geometric entities and the plurality of first geometric entities. The occupancy representation may have a hierarchical structure where the first geometric entities may be on a lower level than the one or more second geometric entities. The system can also detect collision of the object with another object by using the occupancy representation of the object and an occupancy representation of the other object.

Abstract: Systems, apparatuses and methods may provide for technology that adjusts a plurality of weights in a neural network model and adjusts a plurality of activation functions in the neural network model. The technology may also output the neural network model in response to one or more conditions being satisfied by the plurality of weights and the plurality of activation functions. In one example, two or more of the activation functions are different from one another and the activation functions are adjusted on a per neuron basis.

Abstract: This disclosure describes systems, methods, and devices related to reducing noise in and improving speech signals and radio signals. A device may identify a radio frequency signal received by radio frequency circuity at a time; apply a Short Time Fourier Transform (STFT) to the radio frequency signal to generate a STFT signal; identify a signal-to-noise ratio associated with the radio frequency signal at the time; identify a packet-error-rate associated with the radio frequency signal at the time; identify activity when the multiplexed radio frequency signal is received by the radio frequency circuity at the time; select, based on the signal-to-noise ratio, the packet-error-rate, and the activity, a spectral mask to be applied to the STFT signal; apply the selected spectral mask to the STFT signal to generate a clean STFT signal; and apply an inverse STFT to the clean STFT signal to generate a clean radio frequency signal.

Abstract: Systems, apparatuses and methods may provide for technology that selects a fractional derivative value, determines a derivative operation based on the fractional derivative value, applies the derivative operation to an activation function to obtain a deformable fractional filter, generates a mask based on the deformable fractional filter, and convolves the mask with input data.

Abstract: A system, article, and method of detection of acoustic source aiming direction uses multiple microphones.

Abstract: A localization system includes a first sensor, configured to detect a vibration and generate a first electrical signal in response to the detection of the vibration; a second sensor, configured to detect a vibration and generate a second electrical signal in response to the detection of the vibration; a third sensor, configured to detect a vibration and generate a third electrical signal in response to the detection of the vibration; and a processor, configured to determine a position of a source of the vibration based on the first electrical signal, the second electrical signal, and the third electrical signal.

Abstract: An apparatus is provided for deep learning. The apparatus accesses a neural network including an input layer, hidden layers, and an output layer. The apparatus adds an activation function to one or more of the hidden layers of the hidden layers and output layer. The activation function includes a tunable parameter, the value of which can be adjusted during the training of the neural network. The apparatus trains the neural network by inputting training samples into the neural network and determining internal parameters of the neural network based on the training samples. Determining the internal parameters includes determining a value of the tunable parameter based on the training samples. The apparatus may determine two different values of the tunable parameter for two different layers. The activation function may include another tunable parameter. The apparatus can determine a value for the other tunable parameter during the training of the neural network.

Abstract: Methods and apparatus to facilitate autonomous navigation of robotic devices. An example autonomous robot includes a region model analyzer to: analyze a first image of an environment based on a first neural network model, the first image captured by an image sensor of the robot when the robot is in a first region of the environment; and analyze a second image of the environment based on a second neural network model, the second image captured by the image sensor when the robot is in a second region of the environment, the second neural network associated with the second region. The example robot further includes a movement controller to: autonomously control movement of the robot within the first region toward the second region based on the analysis of the first image; and autonomously control movement of the robot within the second region based on the analysis of the second image.

Abstract: An audio processing device and method uses audio signals from a virtual rotating microphone for acoustic angle of arrival detection using a doppler effect technique.

Abstract: An example apparatus including processor circuitry to execute instructions to join a video conferencing meeting including a first audio channel, the first audio channel including first binaural audio at a first angle; in response to a user joining a second audio channel in the video conferencing meeting, select a second angle for the second audio channel and a third angle for the first audio channel; generate a superimposed binaural audio including second binaural audio and third binaural audio; and output the superimposed binaural audio such that first audio data from the first audio channel included in the third binaural audio is to appear to originate from a first position based on the third angle and second audio data from the second audio channel included in the second binaural audio is to appear to originate from a second position based on the second angle.

Abstract: The present disclosure provides a machine learning model where each activation node within the model has an adaptive activation function defined in terms of an input and a hyperparameter of the model. Accordingly, each activation node can have a separate of distinct activation function, based on the adaptive activation function where the hyperparameter for each activation node is trained during overall training of the model. Furthermore, the present disclosure provides that a set of adaptive activation functions can be provided for each activation node such that a spike train of activations can be generated.

Abstract: In some embodiments, a method calculates a first color metric value from a detection window in a first image that is detected by a visual system. A second image of pixels is generated where the pixels include one or more second color metric values that meet the first color metric value within a threshold. The method compares the second image to an object. When the comparison meets a criterion, the method transitions from a first mode to a second mode. When the comparison does not meet the criterion, the method continues to operate in the first mode.

Abstract: Methods, apparatus, systems, and articles of manufacture are disclosed to identify and interpret code. An example apparatus includes a natural language (NL) processor to process NL features to identify a keyword, an entity, and an intent of an NL string included in an input retrieved from a user; a database driver to transmit a query to a database including an ontological representation of a version control system, wherein the query is a parameterized semantic query including the keyword, the entity, and the intent of the NL string; and an application programming interface (API) to present to the user a code snippet determined based on the query, the code snippet being at least one of uncommented or non-self-documented.

Abstract: Methods and apparatus to calibrate a multicamera system based on a human pose are disclosed. An example apparatus includes an object identifier to identify a first set of coordinates defining first locations of anatomical points of a human in a first image captured by a first camera and identify a second set of coordinates defining second locations of the anatomical points in a second image captured by a second camera. The apparatus includes a pose detector to detect, based on at least one of the first or second sets of coordinates, when the human is in a particular pose. The apparatus includes a transformation calculator to, in response to detection of the human in the particular pose, calculate a relative transformation between the first camera and the second camera based on a first subset of the first set of coordinates and a second subset of the second set of coordinates.

Abstract: Apparatuses and methods for transforming a signal from a vibration sensor into an approximation of a signal from a microphone are disclosed. The vibration sensor signal may be transformed by first obtaining a smooth average spectrum and corresponding deviation data for the vibration sensor signal. The smooth average spectrum may be passed through a mapping neural network to obtain a transformed smooth average spectrum that approximates a smooth average spectrum that would be obtained from a microphone. The transformed signal is then obtained by recombining the transformed smooth average spectrum with the deviation data.

Abstract: In some embodiments, a method calculates a first color metric value from a detection window in a first image that is detected by a visual system. A second image of pixels is generated where the pixels include one or more second color metric values that meet the first color metric value within a threshold. The method compares the second image to an object. When the comparison meets a criterion, the method transitions from a first mode to a second mode. When the comparison does not meet the criterion, the method continues to operate in the first mode.

Abstract: Methods and apparatus to facilitate autonomous navigation of robotic devices. An example autonomous robot includes a region model analyzer to: analyze a first image of an environment based on a first neural network model, the first image captured by an image sensor of the robot when the robot is in a first region of the environment; and analyze a second image of the environment based on a second neural network model, the second image captured by the image sensor when the robot is in a second region of the environment, the second neural network associated with the second region. The example robot further includes a movement controller to: autonomously control movement of the robot within the first region toward the second region based on the analysis of the first image; and autonomously control movement of the robot within the second region based on the analysis of the second image.

Abstract: A wearable gesture recognition device is disclosed that provides gesture recognition for gestures that may include a hold or steady-state component, and may account and adapt for real-time fit-level changes. The wearable gesture recognition device may integrate a photoplethysmographic (PPMG) and a piezoelectric (PZE) sensor such that respective sensor signals may be used individually, or in concert for gesture recognition. Thus the wearable gesture recognition device generally disclosed herein may advantageously perform gesture recognition through the fusion of PPMG and PZE signals. To support continuous gesture recognition, the wearable gesture recognition device may use a low-power activity detection scheme that analyzes a PZE signal prior to higher-power gesture classification. Moreover, the wearable gesture recognition device may provide power management by controlling a duty-cycle of the PPMG sensor without necessarily reducing recognition performance.

Abstract: Embodiments of the present disclosure provide techniques and configurations for an apparatus for detection of a change of electromagnetic field in response to a gesture, to identify the gesture that caused the field change. In one instance, the apparatus may include a first conducting component having first features for the disposal on or around a portion of a user's body, to generate an electromagnetic field in response to a receipt of a source signal. The apparatus may further include a second conducting component having second features for the disposal on or around a portion of the user's body, at a distance from the first conducting component, to provide an indication of a change in the electromagnetic field over time, to identify a change of a position of the user's body portion (gesture) that causes the change in the electromagnetic field. Other embodiments may be described and/or claimed.

Abstract: Embodiments of the present disclosure provide techniques and configurations for an apparatus for a user's physiological context measurements. In one instance, the apparatus may include a processing block and first and second piezoelectric sensors coupled with the processing block. The first and second sensors may include respectively first and second electrodes to provide contact with a user's body in response to mounting of the apparatus on the user's body. The processing block may comprise a multi-modal sensing system configured to perform measurements of a user's physiological context during the contact of the user's body with the first and second electrodes, based at least in part on a voltage signal generated by the user's body and provided to the processing block via the first and second electrodes. Other embodiments may be described and/or claimed.