Low amplitude noise filter

We employ a strategy to remove low amplitude noise using a rolling window strategy, which effectively applies a mean filter if and only if the absolute deviation from the median within the window exceeds a certain threshold.

@nb.jit(nopython=True)  
def low_amplitude_filter(  
    value: np.ndarray,  
    window_size: int = 10,  
    stride: int = 1,  
    threshold: float = 1e-6,  
) -> np.ndarray:  
    """  
    Remove small amplitude fluctuations from a time series by    
    replacing the values in a window with the mean of the window if 
    the values in the window are within a certain threshold of the    
    median value.  
    
    :param value: The array to smooth.
    :param window_size: The size of the window to smooth.
    :param stride: The stride between the windows.
    :param threshold: The threshold for the difference between the        
					  values in the window.    
    """
    new_arr = value.copy()  
    for i in range((value.size - window_size) // stride):  
        s = slice(i * stride, i * stride + window_size)  
  
        arr_s = new_arr[s]  
        condition = np.all(np.abs(arr_s - np.median(arr_s)) < threshold)  
        # condition = np.max(arr_s) - np.min(arr_s) < threshold  
        if condition:  
            new_arr[s] = np.mean(arr_s)  
        else:  
            new_arr[s.start : s.start + stride] = arr_s[:stride]  
  
    return new_arr