Abstract: A device-agnostic system for detecting one or more sound profiles, each of the one or more sound profiles including at least one spectrogram produced from sound associated with a touch input and mapped to a control operation for a device, the system including a transducer and a processor. The transducer can detect a sound produced from touching a surface and produce an electrical signal from the detected sound. The processor is configured to receive the electrical signal from the transducer, convert the received electrical signal to a spectrogram, determine, using image recognition, that the spectrogram meets or surpasses a similarity threshold to one of the one or more sound profiles, and change at least one characteristic of the device based on the control operation mapped to the determined sound profile.

Abstract: A headset comprising a controller configured to receive wind information comprising information relating to a direction of wind in a predetermined coordinate system, direction information relating to a direction of the headset in the predetermined coordinate system, and an input audio signal and to generate an output audio signal based on the wind information, the direction information and the input audio signal.

Abstract: A signal conversion from an input signal to an output signal where the filter used is factorized so that the conversion comprises determining 1) only a first factor at each sampling time of the input signal, where this first factor is independent on the sampling times of the output signal, and 2) only a second factor at each sampling time of the output signal, where this second factor is independent of the sampling times of the input signal. This reduces the computational load for this conversion. In addition, for most filters, the factors may be calculated recursively further increasing the computational load and also reducing the storage requirements. This allows for instantaneous changes in the sampling rates or non-uniform sampling rates with low computational requirements and low memory usage.

Abstract: A headset comprising a controller configured to receive wind information comprising information relating to a direction of wind in a predetermined coordinate system, direction information relating to a direction of the headset in the predetermined coordinate system, and an input audio signal and to generate an output audio signal based on the wind information, the direction information and the input audio signal.

Abstract: The present disclosure relates to a method for performing inference on input data using a neural network and a processing device employing the aforementioned method. The method comprises the steps of obtaining and storing input data, obtaining parameter data indicating the parameters of the first layer and storing the parameter data in a parameter data storage location and processing the input data using the first layer parameter data, to form first layer output data. The method further comprises storing the first layer output data, obtaining parameter data of the second layer and storing the second layer parameter data by replacing the first layer parameter data with the second layer parameter data, processing the first layer output data using the stored second layer parameter data to form second layer output data; and storing the second layer output data.

Abstract: A signal conversion from an input signal to an output signal where the filter used is factorized so that the conversion comprises determining 1) only a first factor at each sampling time of the input signal, where this first factor is independent on the sampling times of the output signal, and 2) only a second factor at each sampling time of the output signal, where this second factor is independent of the sampling times of the input signal. This reduces the computational load for this conversion. In addition, for most filters, the factors may be calculated recursively further increasing the computational load and also reducing the storage requirements. This allows for instantaneous changes in the sampling rates or non-uniform sampling rates with low computational requirements and low memory usage.

Abstract: The present disclosure shows new mechanisms for sampling an input signal. In particular, some embodiments of the present disclosure include a new type of a level-crossing sampling mechanism called a derivative level-crossing sampling (D-LCS). At a high level, D-LCS involves quantizing the derivative of an input signal when the derivative of the input signal crosses one of the quantization thresholds. For certain class of signals, the derivative of the input signal can vary at a slower speed compared to the amplitude of the input signal. Therefore, by sampling the derivative of the input signal, instead of the input signal itself, the number of samples per unit time can be reduced.

Abstract: The derivative of an input signal is level-crossing-sampled and the resulting samples are transmitted. At the receiver, the resulting samples are fed to a zero-order hold followed by an integrator, which results in piecewise-linear reconstruction. The systems and methods are further refined to reduce the number of samples generated per unit of time, compared to methods based on zero-order-hold reconstruction, for a given signal-to-error ratio, without significant hardware overhead.

Abstract: Systems and methods for level-crossing sampling and reconstruction technique are disclosed. The derivative of the input signal is level-crossing-sampled, and the resulting samples are transmitted. At the receiver, these samples are fed to a zero-order hold followed by an integrator, which results in piecewise-linear reconstruction. The disclosed systems and methods are further refined to reduce the number of samples generated per unit of time, compared to methods based on zero-order-hold reconstruction, for a given signal-to-error ratio, without significant hardware overhead.