Abstract: Systems and methods for audio privacy in network video surveillance systems are described. A video camera may include an image sensor and a microphone to generate a video stream. Responsive to detecting a human speaking condition in the video stream, the audio data may be selectively modified to mask a human voice component of the audio data for storing and/or displaying the surveillance video stream.

Abstract: In one embodiment, a sensor device includes a sensor configured to obtain sensor data and a communication interface configured to communicate with a set of secondary sensor devices. The set of secondary sensor devices comprise a set of secondary sensors and a set of secondary processing devices. The set of secondary sensors obtain additional sensor data. The set of secondary processing devices process the additional sensor data based on a set of secondary machine learning models, and generate activation data based on the additional sensor data. The sensor device also includes a processing device. The processing device is also configured to receive the activation data from the set of secondary sensor devices and cause the sensor to obtain first sensor data based on the activation data. The processing device is further configured to generate one or more inferences based on the first sensor data and a machine learning model.

Abstract: A surveillance system includes at least one image or video capture device and a controller configured to determine a change in location for the at least one image or video capture device from a first location to a second location. First image or video data is received from the at least one image or video capture device at the second location, and in response to the location change, a preconfigured neural network is obtained or weights for a neural network are obtained based at least in part on the received first image or video data. Second image or video data is received from the at least one image or video capture device at the second location and an inference operation is performed on the second image or video data by processing the second image or video data using the obtained weights or the obtained preconfigured neural network.

Abstract: This disclosure relates to an apparatus for object detection. The apparatus comprises a video camera, an object detector, and a controller. The video camera may be configured to generate a video stream of frames. The object detector may be configured to accept the video stream as input data and to perform object detection. The controller may be coupled to the video camera and the object detector. The controller may be configured to manage object detection in order to satisfy a performance metric and/or operate within an operational constraint.

Abstract: A device configured for processing time-series data within an asynchronous neural network may include a processor configured to execute the neural network. The device may further include a multi-step convolution pathway wherein the output of at least one step includes one or more feature maps. Additionally, a multi-step upsampling pathway with steps having corresponding convolution step inputs is included. The device further utilizes feature map data from at least one step of the multi-step convolution process as input data in at least one corresponding step of the upsampling process. The device also includes an inference frequency controller to receive input data and transmit a processing frequency signal to the neural network. The neural network can then generate feature maps at a reduced frequency within the multi-step convolution pathway, and utilize previously processed feature maps as input data within the multi-step upsampling pathway until a subsequent feature map is generated.

Abstract: In one embodiment, a sensor device includes a sensor configured to obtain sensor data and a communication interface configured to communicate with a set of secondary sensor devices. The set of secondary sensor devices comprise a set of secondary sensors and a set of secondary processing devices. The set of secondary sensors obtain additional sensor data. The set of secondary processing devices process the additional sensor data based on a set of secondary machine learning models, and generate activation data based on the additional sensor data. The sensor device also includes a processing device. The processing device is also configured to receive the activation data from the set of secondary sensor devices and cause the sensor to obtain first sensor data based on the activation data. The processing device is further configured to generate one or more inferences based on the first sensor data and a machine learning model.

Abstract: Remote neural network retraining for surveillance systems is disclosed. In some cases, systems and methods enable remote training and retraining of neural networks, while processing data in real life using the trained and retrained neural networks locally at the surveillance system. The surveillance system can determine a change of location and can retrain the neural network and/or initiate the retraining of the neural network remotely, such as on a cloud server, to retrain the neural network based on new image and/or video data taken from the new location. The remote server can transmit the retrained neural network and/or weights for nodes of the retrained neural network back to the surveillance system. The surveillance system can then update its neural network and process future image and/or video data based on the retrained neural network and/or weights.

Abstract: An example system and method may provide for important data files to be stored in differently reliable storage medium based upon an importance score that is representative of the importance of the data file. A data file importance score is generated based on the data file. The data file is stored in a first data storage zone of a storage medium in a data storage device, in response to the data file importance score being greater than a data file importance threshold. The data file is stored in a second data storage zone of the storage medium in the data storage device, in response to the data file importance score being less than the data file importance threshold.

Abstract: An example system and method may provide an importance score for a data file based on the content of the data file. An importance score may be used to determine whether to store the data file in a regular reliability storage media or in a higher reliability storage media. A controller generates a document vector for a data file based on content processed from a data file. The data file includes metadata and the content. The controller generates, using an artificial intelligence (AI) model and the document vector, a data file importance score for the data file. The controller then stores the data file in one of the first data storage zone and the second data storage zone based on the data file importance score.

Abstract: A data storage device is disclosed comprising a non-volatile storage medium (NVSM), and a head configured to access the NVSM. During a first interval, the head is used to write first data to a first segment of the NVSM, and during a second interval, the head is used to read the first data from the first segment of the NVSM and erase at least part of the first data from the first segment of the NVSM.

Abstract: In one embodiment, a sensor device includes a sensor configured to obtain sensor data and a communication interface configured to communicate with a set of secondary sensor devices. The set of secondary sensor devices comprise a set of secondary sensors and a set of secondary processing devices. The set of secondary sensors obtain additional sensor data. The set of secondary processing devices process the additional sensor data based on a set of secondary machine learning models, and generate activation data based on the additional sensor data. The sensor device also includes a processing device. The processing device is also configured to receive the activation data from the set of secondary sensor devices and cause the sensor to obtain first sensor data based on the activation data. The processing device is further configured to generate one or more inferences based on the first sensor data and a machine learning model.

Abstract: This disclosure relates to an apparatus for object detection. The apparatus comprises a video camera, an object detector, and a controller. The video camera may be configured to generate a video stream of frames. The object detector may be configured to accept the video stream as input data and to perform object detection. The controller may be coupled to the video camera and the object detector. The controller may be configured to manage object detection in order to satisfy a performance metric and/or operate within an operational constraint.

Abstract: A semiconductor device is disclosed including one or more integrated memory modules. Each integrated memory module may include a pair of semiconductor die, which together, operate as a single, integrated memory. In one example, the first die may include the memory cell array and the second die may include the logic circuit such as CMOS integrated circuits. In one example, the first die may be flip-chip bonded to the second die.

Abstract: Remote neural network retraining for surveillance systems is disclosed. In some cases, systems and methods enable remote training and retraining of neural networks, while processing data in real life using the trained and retrained neural networks locally at the surveillance system. The surveillance system can determine a change of location and can retrain the neural network and/or initiate the retraining of the neural network remotely, such as on a cloud server, to retrain the neural network based on new image and/or video data taken from the new location. The remote server can transmit the retrained neural network and/or weights for nodes of the retrained neural network back to the surveillance system. The surveillance system can then update its neural network and process future image and/or video data based on the retrained neural network and/or weights.

Abstract: A data storage system may include multiple data storage devices, such as solid-state drives, an enclosure housing the devices, and a thermal bridge positioned in a gap between and in contact with each of the enclosure and a device, where the enclosure is cooler than the device. Thus, heat is conducted away from a hot spot of the device and to the enclosure. The thermal bridge may be flexible enough to bridge different sized gaps, while stiff enough to generate contact forces applied to the enclosure and the device. For example, the thermal bridge may be constructed primarily of copper and configured to function like a compression spring.

Abstract: A semiconductor device is disclosed including semiconductor dies stacked with an offset in two orthogonal directions. TSVs may then be formed connecting corresponding die bond pads on respective dies in the stack. By offsetting the dies in two orthogonal directions, the overall stepped offset, and consequently the size of the unused keep-out area of the stack, is reduced.

Abstract: A semiconductor device is disclosed including semiconductor dies stacked with an offset in two orthogonal directions. TSVs may then be formed connecting corresponding die bond pads on respective dies in the stack. By offsetting the dies in two orthogonal directions, the overall stepped offset, and consequently the size of the unused keep-out area of the stack, is reduced.

Abstract: A compact back-flow stopper using an elastic flexure mechanism for a cooling fan is shown and described. The backflow stopper assembly comprises a frame configured to structurally support a fin array when coupled to a fan. The fin array comprises a plurality of flexural deformation elements and associated fin elements arrayed in a radial arrangement to establish a pathway for airflow, each of the flexural deformation elements configured to move an attached fin element responsive to airflow impacting the attached fin element. Flex limiter elements are coupled to the frame and configured to limit flexure of the fin elements beyond a predetermined flexure in relation to the frame to stop backflow of air through the fin array. A fan assembly with a backflow stopper is shown and described. Also, a data storage assembly using fan assemblies with backflow stoppers is also shown and described.

Abstract: A data storage device is disclosed comprising a first head actuated over a first disk surface, wherein the first head comprises a write coil and a write-assist component. First data is written to the first disk surface using the write coil and the write-assist component of the first head. An electrical bias is applied to the write-assist component of the first head to physically damage the write-assist component, thereby rendering the first disk surface read-only.

Abstract: A data storage system may include multiple data storage devices, such as hard disk drives, a cooling fan, and a sound balancer structure between the fan and the drives. The sound balancer is intentionally positioned where it substantially balances, about the center of rotation of and along at least one axis of a fan-facing storage device, the sound pressure from the fan that impinges upon the fan-facing face of the storage device. This redistribution of the sound pressure suppresses sound pressure-induced torque upon and therefore rotational vibration of the storage device, which in turn enhances the track following capability of the storage device.