Noise suppression device and noise suppression method

A noise suppression method that estimates desired information from only observed information in which noise is mixed with the desired information, the noise suppression method comprising:

• a correlation computation step, performed by a correlation computation section configured for calculating a correlation matrix (Pp1(n+1|n)), resp. a correlation value (Pp2(n+1|n)[1,1]) of an estimation error at time n+1, when a system state vector (xp1(n+1),xp2(n+1)) at time n+1, including the desired information (d(n+1)), is estimated based on observed information up to time n;
• a weighting coefficient calculation step, performed by a weighting coefficient calculation section configured for using the correlation matrix, resp. the correlation value to calculate a weighting vector (kp1(n+1)), resp. a weighting coefficient (kp2(n+1)[1]), for specifying a relationship between an optimum estimate of the state vector at time n+1 based on observed information up to time n+1 ({circumflex over (x)}p1(n+1|n+1)), an optimum estimate of the state vector at time n+1 based on observed information up to time n ({circumflex over (x)}p1(n+1|n)), and an estimation error of an observed information at time n+1 only, resp. a relationship between an optimum estimate of the state vector at time n+1 based on observed information up to time n+1 ({circumflex over (x)}p2(n+1|n+1)) and observed information at time n+1 only (yp2(n+1));
• and
• an optimum estimate calculation step, performed by an optimum estimate calculation section configured for using the relationship, to calculate an optimum estimate of the state vector at time n+1 based on observed information up to time n+1 ({circumflex over (x)}p1(n+1|n+1),{circumflex over (x)}p2(n+1|n+1)), wherein
• the noise suppression method uses a Kalman filter based on a system state model having the desired information (d(n+1)) as driving source.

A non-transitory computer-readable storage medium having stored thereon a noise suppression program for estimating desired information from only observed information in which noise is mixed with the desired information, the noise suppression program causing a computer to execute:

• a correlation computation step, performed by a correlation computation section configured for calculating a correlation matrix Pp1(n+1|n)), resp. a correlation value (Pp2(n+1|n)[1,1]) of an estimation error at time n+1, when a system state vector (xp1(n+1),xp2(n+1) at time n+1, including the desired information (d(n+1)), is estimated based on observed information up to time n;
• a weighting coefficient calculation step, performed by a weighting coefficient calculation section configured for using the correlation matrix, resp. the correlation value to calculate a weighting vector (kp1(n+1)), resp. a weighting coefficient (kp2(n+1)[1]), for specifying a relationship between an optimum estimate of the state vector at time n+1 based on observed information up to time n+1 ({circumflex over (x)}p1(n+1|n+1)), an optimum estimate of the state vector at time n+1 based on observed information up to time n ({circumflex over (x)}p1(n+1|n)), and an estimation error of an observed information at time n+1 only, resp. a relationship between an optimum estimate of the state vector at time n+1 based on observed information up to time n+1 ({circumflex over (x)}p2(n+1|n+1)) and observed information at time n+1 only (yp2(n+1));
• and
• an optimum estimate calculation step, performed by an optimum estimate calculation sections configured for using the relationship, to calculate an optimum estimate of the state vector at time n+1 based on observed information up to time n+1 ({circumflex over (x)}p1(n+1|n+1),{circumflex over (x)}p2(n+1|n+1)), wherein
• the noise suppression program causes the computer to execute a noise suppression method using a Kalman filter based on a system state model having the desired information (d(n+1)) as driving source.