Signal processing apparatus, signal processing method, and program using an optimum configuration determined by machine learning to output an image in accordance with an expectation

A signal processing method in which a signal processing apparatus executes processing of:

• accepting input of image data and data from an inertial measurement unit (IMU);
• performing sensor fusion processing for obtaining consistencies between the image data and the data from the IMU;
• outputting an output value based on the input image data;
• calculating a difference between an expectation based on a desired performance of a stabilization function to be applied to the input image data and the output value;
• searching for an optimum configuration by identifying, at least one of a physical sensor or an image recognition algorithm, by performing machine learning using information on the difference and information based on the image data to approximate the output value to the expectation;
• and
• subsequently reconfiguring a signal processing algorithm of the signal processing apparatus based on the identified optimum configuration and processing subsequent image data using the reconfigured signal processing algorithm to generate an output image that approximates the expectation.

A non-transitory computer readable storage device storing a program causing a signal processing apparatus to execute processing of:

• accepting input of image data and data from an inertial measurement unit (IMU);
• performing sensor fusion processing for obtaining consistencies between the image data and the data from the IMU;
• outputting an output value based on the input image data;
• calculating a difference between an expectation based on a desired performance of a stabilization function to be applied to the input image data and the output value;
• searching for an optimum configuration by identifying, at least one of a physical sensor or an image recognition algorithm, by performing machine learning using information on the difference and information based on the image data to approximate the output value to the expectation;
• and
• subsequently reconfiguring a signal processing algorithm of the signal processing apparatus based on the identified optimum configuration and processing subsequent image data using the reconfigured signal processing algorithm to generate an output image that approximates the expectation.