| 摘要 |
A stochastic Bayesian non-linear filtering system and method that improves the filtering of noisy signals by providing efficiency, power, speed, and flexibility. The filter only requires the likelihood function p(observation|state) to determine the system state and works in various measurement models. This allows for the processing of noisy signals to be used in real time, such as in a biofeedback device that senses noisy surface electromyography muscle electrical activity, filters the sensed signal using the nonlinear filtering method, and provides vibrations based on the muscular activity. |
| 主权项 |
1. A method for generating muscle activity feedback using a nonlinear filter to estimate a hidden state of an environment, wherein the hidden state comprises a plurality of state values, designated x, that characterize the environment over time; the method comprising:
receiving a noisy signal from a sensor component comprising electrodes; wherein the noisy signal represents electrical activity produced by a user's muscle; wherein the noisy signal comprises a plurality of time-varying signal measurements, designated z; determining an observation model, designated p(z|x), that relates a probability of signal measurements to the plurality of state values; determining a system dynamics model, designated F, that relates the hidden state to a change in state values based on dynamics of the environment; determining an initial estimate of a probability density function that characterizes a probability of one of said state values at an initial time; repeatedly updating a current estimate of the probability density function, designated p(x,t), that characterizes a probability of one of said state values based on a time, designated t; wherein the updated probability density function is based, at least in part, on:
a product of the system dynamics model and the current probability density function; anda product of the current probability density function and a logarithm of the observation model; outputting a maximum likelihood estimate based on the current estimate of the probability density function, wherein the maximum likelihood estimate represents an estimate of a state value at each time; generating an estimate of the plurality of state values, x, based on the maximum likelihood estimate at each time; and applying muscle activity feedback to the user based on the estimate of the plurality of state values, x, at each time; wherein the feedback is indicative of electrical activity produced by the user's muscles; wherein continuous-time nonlinear filtering of the noisy signal is acquired based on the observation model, p(z|x), and signal measurements, z. |
