A general least mean square interference technique is provided for effective adaptive filtering. The gradient adaptive learning rate methodology can now handle non-stationary data with the Interference normalised least mean square technique. Because of issues like duplicate talk and echo route variance, echo cancellation is made more difficult because the learning rate must be adjusted. Frequency domain echo cancelers learn at different rates, which can be altered in a novel fashion. Normalized least mean square method normalised learning rate under noise is used to calculate an optimal learning rate. This double-talk detection technique exceeds the competition while also being incredibly simple to implement. A number of least mean square (LMS)-type algorithms have been investigated in place of their recursive equivalents of IVM or TLS/DLS, which involve large calculations. As a result of these findings, we provide a consistent LMS type technique for the data least squares estimate problem. This unique approach normalizes step size and estimates the variance of the noise in a heuristic manner using the geometry of the mean squared error function, resulting in rapid convergence and robustness against environmental noise.