mnru

Modulated noise reference unit (MNRU)

Since R2026a

collapse all in page

Syntax

audioOut = mnru(audioIn,Fs)

audioOut = mnru(___,Name=Value)

[audioOut,audioPT] = mnru(___)

[audioOut,audioPT,noiseSig] = mnru(___)

mnru(___)

Description

audioOut = mnru(audioIn,Fs) returns the audio signal audioOut with noise added according to the algorithms defined in the ITU-T Recommendation P.810 (03/2023) standard [1]. audioIn is the real-valued input audio signal, and fs is the sampling rate in hertz. The function assumes a signal-to-modulated-noise ratio of 15 dB by default.

audioOut = mnru(___,Name=Value) specifies options using one or more name-value arguments.

[audioOut,audioPT] = mnru(___) additionally outputs the filtered signal without noise (pass-through signal), audioPT.

[audioOut,audioPT,noiseSig] = mnru(___) also outputs the noise signal, noiseSig.

example

mnru(___) opens the Audio Viewer app for signal playback and visualization.

Examples

collapse all

Generate MNRU

Open Live Script

Define a signal-to-modulated-noise ratio of 10 dB and an ITU-recommended reference level of –26.15 dBov.

QdB = 10;
refLevel = -26.15;

Read the audio file and generate the MNRU of the audio file according to the signal-to-modulated-noise ratio and the reference level.

[x,fs] = audioread("SpeechDFT-16-8-mono-5secs.wav");
[y,xf] = mnru(x,fs,NoiseRatio=QdB,ReferenceLevel=refLevel);

Listen to the clean reference signal, followed by the MNRU signal.

as = audiostreamer(SampleRate=fs);
play(as,xf)
play(as,y)

Input Arguments

collapse all

`audioIn` — Input signal
real-valued 2-D matrix

Input signal, specified as a real-valued 2-D matrix. The function treats the columns of the matrix as audio channels.

Data Types: single | double

`Fs` — Sample rate (Hz)
positive scalar equal to or greater than 8000

Sample rate of the input signal, specified as a positive scalar equal to or greater than 8000. Units are in Hz.

Data Types: single | double

Name-Value Arguments

collapse all

Specify optional pairs of arguments as Name1=Value1,...,NameN=ValueN, where Name is the argument name and Value is the corresponding value. Name-value arguments must appear after other arguments, but the order of the pairs does not matter.

Example: mnru(X,FS,NoiseRatio=5,DCFilter=true,NoiseFilter=false,OutputFilter=true,ReferenceLevel=-26.15);

`NoiseRatio` — Signal-to-modulated-noise ratio (dB)
15 (default) | real-valued scalar

Signal-to-modulated-noise ratio, specified as a real-valued scalar. Typical values are 5, 15, 25, and 35. Units are in dB.

Example: 5

Data Types: single | double

`ReferenceLevel` — Reference level (dBov)
nonpositive scalar

Reference level, specified as a nonpositive scalar. The mnru function normalizes the output signals so that the level of the pass-through signal, audioPT, matches this reference level. Units are in dBov. The ITU-T standard recommends a reference level of –26.15 dBov to avoid clipping. If unspecified, the function does not normalize the output signal.

Example: -26.15

Data Types: single | double

`DCFilter` — Option to enable DC blocking filter
`true` | `false`

Option to enable a DC blocking filter to input signal, specified as true or false. This filter removes DC components in the input audio signal and prevents the addition of extra noise to the output signal. If you do not specify this argument, the function applies the filter only if the Fs input argument is equal to or smaller than 16 kHz.

Data Types: logical

`OutputFilter` — Option to enable lowpass filter
`true` | `false`

Option to enable a lowpass filter to output signal, specified as true or false. If you do not specify this argument, the function applies the filter only if the Fs input argument is equal to or smaller than 16 kHz.

Data Types: logical

`NoiseFilter` — Option to enable long-term average speech spectrum filter
`true` | `false`

Option to enable a long-term average speech spectrum (LTASS) filter to the Gaussian noise source, specified as true or false. This filter is compliant with ITU-T Recommendation P.50. If unspecified, the filter is applied only if the Fs input argument is greater than 16 kHz.

Data Types: logical

Output Arguments

collapse all

`audioOut` — Output signal with added noise
real-valued 2-D matrix

Output signal with added noise, returned as a real-valued 2-D matrix. The size of this matrix matches the size of the audioIn input argument.

Data Types: single | double

`audioPT` — Filtered signal without added noise
real-valued 2-D matrix

Filtered signal without added noise (pass through), returned as a real-valued 2-D matrix. The size of this matrix matches the size of the audioIn input argument.

Data Types: single | double

`noiseSig` — Noise signal
real-valued 2-D matrix

Noise signal, returned as a real-valued 2-D matrix. The size of this matrix matches the size of the audioIn input argument.

Data Types: single | double

Algorithms

collapse all

MNRU system overview

The Modulated Noise Reference Unit (MNRU) is a signal-correlated noise generator designed to introduce controlled degradations to speech signals. Defined by the ITU-T Recommendation P.810 (03/2023) standard [1] (hereafter referred to as the standard), the MNRU includes the previous P.810 (1996) version, which supports bandwidth up to 16 kHz, and the ITU-T Recommendation P.50, which supports full-bandwidth signals up to 20 kHz. This technology is widely used for testing and evaluating speech communication systems under various conditions, generally as reference conditions (anchors) in subjective evaluations such as MOS (ITU-T Rec. P.800.1) tests.

The MNRU system comprises several components: a Gaussian noise source generating white noise, an optional long-term average speech spectrum (LTASS) filter that shapes the noise, an optional DC blocking filter, and an optional lowpass output filter.

This table shows the default values of the optional filters (NoiseFilter, DCFilter, and OutputFilter) based on the sampling rate Fs.

Optional Filters	Default Status When 8 kHz ≤ Sampling rate `Fs` ≤ 16 kHz	Default Status When Sampling rate `Fs` > 16 kHz
LTASS filter	off	on
DC blocking filter	on	off
Output filter	on	off

Degradation Algorithm

To produce a specific MNRU condition or a certain level of distortion in the audio, the system uses Q, which is the ratio of the clean signal to the distorted signal. You can set this ratio using the NoiseRatio argument. Degradation is defined as

$y (k) = x (k) [1 + 10^{\frac{- Q}{20}} N (k)],$

where k corresponds to the sample index, x(k) is the input speech sample, N(k) is the Gaussian noise (white noise for a low sample rate or LTASS-shaped noise for a high sample rate), Q is the ratio in dB, and y(k) is the degraded output sample. The noise signal N(k) should be a random Gaussian signal. Moreover, the function can clip any peaks 12 dB above RMS level can be clipped, or it can truncate a Gaussian generator with a variance and power of 1 to ±4 (12.0412 dB). The LTASS filter used to shape the Gaussian noise conforms to the Annex A of the standard.

If you enable the optional DC blocking filter and output filter, the output signal is defined as

$y (k) = G {F {x (k)} [1 + 10^{\frac{- Q}{20}} N (k)]},$

where the DC blocking filter F and the frequency response of the output filter G are defined in the standard for both analog and digital implementations.

References

[1] International Telecommunication Union. Modulated noise reference unit (MNRU). ITU-T Recommendation P.810. March, 2003.

mnru

Syntax

Description

Examples

Generate MNRU

Input Arguments

`audioIn` — Input signal
real-valued 2-D matrix

`Fs` — Sample rate (Hz)
positive scalar equal to or greater than 8000

Name-Value Arguments

`NoiseRatio` — Signal-to-modulated-noise ratio (dB)
15 (default) | real-valued scalar

`ReferenceLevel` — Reference level (dBov)
nonpositive scalar

`DCFilter` — Option to enable DC blocking filter
`true` | `false`

`OutputFilter` — Option to enable lowpass filter
`true` | `false`

`NoiseFilter` — Option to enable long-term average speech spectrum filter
`true` | `false`

Output Arguments

`audioOut` — Output signal with added noise
real-valued 2-D matrix

`audioPT` — Filtered signal without added noise
real-valued 2-D matrix

`noiseSig` — Noise signal
real-valued 2-D matrix

Algorithms

MNRU system overview

Degradation Algorithm

References

Extended Capabilities

C/C++ Code Generation
Generate C and C++ code using MATLAB® Coder™.

Version History

See Also

mnru

Syntax

Description

Examples

Generate MNRU

Input Arguments

audioIn — Input signal real-valued 2-D matrix

Fs — Sample rate (Hz) positive scalar equal to or greater than 8000

Name-Value Arguments

NoiseRatio — Signal-to-modulated-noise ratio (dB) 15 (default) | real-valued scalar

ReferenceLevel — Reference level (dBov) nonpositive scalar

DCFilter — Option to enable DC blocking filter true | false

OutputFilter — Option to enable lowpass filter true | false

NoiseFilter — Option to enable long-term average speech spectrum filter true | false

Output Arguments

audioOut — Output signal with added noise real-valued 2-D matrix

audioPT — Filtered signal without added noise real-valued 2-D matrix

noiseSig — Noise signal real-valued 2-D matrix

Algorithms

MNRU system overview

Degradation Algorithm

References

Extended Capabilities

C/C++ Code Generation Generate C and C++ code using MATLAB® Coder™.

Version History

See Also

`audioIn` — Input signal
real-valued 2-D matrix

`Fs` — Sample rate (Hz)
positive scalar equal to or greater than 8000

`NoiseRatio` — Signal-to-modulated-noise ratio (dB)
15 (default) | real-valued scalar

`ReferenceLevel` — Reference level (dBov)
nonpositive scalar

`DCFilter` — Option to enable DC blocking filter
`true` | `false`

`OutputFilter` — Option to enable lowpass filter
`true` | `false`

`NoiseFilter` — Option to enable long-term average speech spectrum filter
`true` | `false`

`audioOut` — Output signal with added noise
real-valued 2-D matrix

`audioPT` — Filtered signal without added noise
real-valued 2-D matrix

`noiseSig` — Noise signal
real-valued 2-D matrix

C/C++ Code Generation
Generate C and C++ code using MATLAB® Coder™.