Method and apparatus for extracting a component signal from a composite signal

A method for generating a composite signal and extracting a portion thereof generated by a predetermined source using a separate composite sensor and a separate reference sensor in a motor vehicle;

• the method comprising the steps of:
• generating a reference signal from the reference sensor relating to activity of a component source;
• generating a composite signal from the composite sensor;
• transmitting the reference signal to an adaptive filter;
• transmitting the composite signal to the adaptive filter;
• outputting the portion of the composite signal based on the reference signal to be analyzed;
• adapting the adaptive filter to remove the portion of the composite signal based on the reference signal by generating an error signal according to the equation:
• e(n)=d(n)+N(n)−{X(n)}T{W(n−1)} wherein d(n) is a portion of the composite signal and minimizing the mean square of the error signal and updating the adaptive filter according to the equation:
• {W(n)}={W(n−1)}+2 μe(n){X(n)}
• wherein μ is an adaptation coefficient used for speed of convergence and stability;
• and
• stabilizing the step of adapting the adaptive filter to minimize the step of adapting including normalizing the step size taken for each iteration of the adaptive filter by generating a normalization coefficient according to the equation:
• μ ⁡ ( n ) = μ 0 { X ⁡ ( n ) } T ⁢ { X ⁡ ( n ) } ′ , ⁢ μ 0 < 1

An extracting assembly for extracting a component signal generated by a component source from a composite signal of a motor vehicle, said extracting assembly comprising:

• at least one accelerometer for generating a reference signal relating to activity of the component source;
• a microphone for sensing the composite signal;
• an adaptive filter operatively connected to said at least one accelerometer and said microphone, said adaptive filter filtering the component signal from the composite signal as a function of the reference signal generated by said reference sensor, said adaptive filter utilizing a least-mean-square algorithm to adapt coefficients of said adaptive filter, said adaptive filter being adaptable by an adaptation step having an initial estimated size;
• said adaptive filter being updated according to the equation:
• {W(n)}={W(n−1)}+2 μe(n){X(n)}
• wherein μ is an adaptation coefficient used for speed of convergence and stability;
• a performance feedback adaptor operatively connected to said adaptive filter and said microphone to generate an error signal which is the addition of the portion of the composite signal associated with the reference signal and the composite signal and the uncorrelated content of the composite signal, said error signal is calculated according to the equation:
• e(n)=d(n)+N(n)−{X(n)}T{W(n−1)} wherein d(n) is a portion of the composite signal correlated to the activity of the component source;
• a normalizer operatively connected to said adaptive filter for normalizing said predetermined size of said adaptation step of said adaptive filter and having a stability factor to generate the stability of said adaptive filter, said normalizer operates on said stability factor according to the equation:
• μ ⁡ ( n ) = μ 0 { X ⁡ ( n ) } T ⁢ { X ⁡ ( n ) } ′ , ⁢ μ 0 < 1 ;
• and
• an output port operatively connected to said adaptive filter for outputting all of the component signal filtered from the composite signal such that the component signal may be analyzed.