39
DSP 07
Noise Reduction
There are two methods available for noise reduction: NR1 and NR2. You can select the method that is more effective
according to the operation mode and reception conditions.
For NR1, the algorithm that functions varies with the operation mode. In voice modes (SSB, FM and AM), noise
reduction featuring the spectral subtraction method specialized for audio signals will function, and in non-voice modes
(CW, FSK and PSK), noise reduction featuring the line enhancer method using the adaptive filter which emphasizes the
periodic signals will function. The functioning algorithm will switch automatically with the mode.
NR2 employs what is known as a SPAC (Speech Processing system using the autocorrelation function) to piece
together only the periodic waveforms detected from the received signal and output them as the received audio signal.
Table 7 Reception Modes and Noise Reduction Algorithms Used
Noise Reduction Method
Receive Mode
SSB, AM, FM
CW, FSK, PSK
NR1
Spectral subtraction
Line enhancer
NR2
SPAC
SPAC
●
NR1 (Spectral Subtraction Method)
NR1 features a spectral subtraction method that estimates noise components contained in the received signal and
eliminates (subtracts) only the estimated noise components from the received signal to highlight the desired signal.
Fig. 46 Conceptual Scheme of NR1 Based on Spectral Subtraction
The spectral subtraction method is developed, intending to improve the intelligibility of weak signals received in SSB
mode. Compared with the conventional NR1 (line enhancer method), the new NR1 has less impact on high-frequency
voice components and realizes audio signal output with noise attenuation and minimum degradation of sound quality.
The TS-890S employs a technique to reduce musical noise (tonal “blip blip” sound that is minutely segmented) that
is inherently generated as a result of the spectral subtraction process. With this technique, musical noise is largely
reduced, and influence by digital processing that is incidental to noise reduction is lowered.
Additionally, with the spectral subtraction method, the NR1 allows adjustments to achieve a smooth noise elimination
effect by controlling the amount of noise attenuation.
In principle, the noise estimation processing of the spectral subtraction method of NR1 also attenuates beats and CW
signals as it identifies constant tones as noise components. Thus, while the conventional NR1 (line enhancer method)
functions to emphasize beats and CW signals, the NR1 with the spectral subtraction method attenuates beats and
CW signals together with noise components. However, since the NR1 featuring the spectral subtraction method is not
intended for eliminating CW and beat signals, it does not have a large attenuation width for such signals.
To eliminate beats and CW signals, use the beat canceler.
The graphs below show how audio signal buried in the noise is highlighted by the NR1 with the spectral subtraction
method.
