S E T U P F O R M C H A P T E R 8
Noise Reduction
Figure 82: Noise Reduction Controls
Noise Reduction (NR) attempts automatic computation of a filter that maximizes the coherent or non-noise
like signals and as a result, filters out the rest of the signal, which includes noise. It is best used for speech
signals with a good signal to noise ratio or tones.
#Taps: This determines the length of the computed filter. The longer the filter, the better the non-
coherent signals (noise) will be cancelled. It also introduces latency equal to the number of filter
taps divided by the sample rate in samples per second and is in addition to Delay (see below). The
larger the number of taps, the longer it takes for the filter to converge but upon achieving
convergence, the better the filter will be.
Delay: Determines how far back to look in the signal before beginning to compute a coherent signal
enhancement filter. With large delays, there is a higher likelihood of detrimental affects to normal
speech. Latency is also introduced that is equal to the Delay.
Gain: Determines the adaptation rate of the filter. The larger the number, the faster the filter will
converge but the less stable it will be.
Block LMS: We have implemented a fast Block LMS (least means square) routine. This routine is
faster than the original routine, gives more taps in the filter, and will work on I/Q signals, which
enables BIN (binaural audio) to remain available.
Automatic Notch Filter
Figure 83: Automatic Notch Filter Controls
85 FlexRadio Systems
To Next Page
Loading...
Need help?
General
