Filter bank approach to adaptive filtering method using independent component analysis

A filter bank approach to default signal separation method using independent component analysis for a feedback or feedforward structures with an arbitrary number of input and output terminals or the structures where some part of adaptive filter coefficients are omitted, comprising the steps of:

• separating each input signal into a plurality of frequency bands after passing the input signal through an analysis filter bank;
• constructing signals whose number is equal to the total number of analysis filters by selecting one sample from an arbitrarily determined number of samples for a filter output corresponding to each frequency band;
• applying the default signal separation method using independent component analysis to said above signals as a whole;
• inserting a number of samples which is one less than said determined number of samples and having zero values between two consecutive samples;
• and
• restoring result signals after passing through a synthesis filter bank.

A filter bank approach to adaptive noise filter method using independent component analysis wherein each of the signals that forms the primary signal, mixed signal of noises and noise signal that forms the reference input is inputted, comprising the steps of:

• separating each input signal into a plurality of frequency bands after passing the input signal through an analysis filter bank;
• constructing signals whose number is equal to the total number of analysis filters by selecting one sample from an arbitrarily determined number of samples for a filter output corresponding to each frequency band;
• applying the adaptive noise filter method using independent component analysis to said above signals as a whole;
• inserting a number of samples which is one less than said determined number of samples and having zero values between two consecutive samples;
• and
• restoring result signals after passing through a synthesis filter bank.

A filter bank approach to default signal separation and adaptive noise filter methods using independent component analysis wherein each of the signals that forms the primary signal, mixed signal of noises and noise signal that forms the reference input is inputted, comprising the steps of:

• separating each input signal into a plurality of frequency bands after passing the input signal through an analysis filter bank;
• constructing signals whose number is equal to the total number of analysis filters by selecting one sample from an arbitrarily determined number of samples for a filter output corresponding to each frequency band;
• applying default signal separation and adaptive noise filter methods using independent component analysis to said above signals as a whole;
• inserting a number of samples which is one less than said determined number of samples and having zero values between two consecutive samples;
• and
• restoring result signals after passing through a synthesis filter bank.