acousticToneToNoiseRatio

Identify and quantify tones in acoustic signals

Since R2023b

Syntax

[tnr,tnrFreq] = acousticToneToNoiseRatio(audioIn,fs)

[tnr,tnrFreq,isProminent] = acousticToneToNoiseRatio(audioIn,fs)

[tnr,tnrFreq,isProminent,timestamps] = acousticToneToNoiseRatio(audioIn,fs,TimeVarying=true)

[___] = acousticToneToNoiseRatio(___,Name=Value)

acousticToneToNoiseRatio(___)

Description

[tnr,tnrFreq] = acousticToneToNoiseRatio(audioIn,fs) detects tones in the audio and returns the corresponding tone-to-noise ratios (TNR) according to ECMA-418-1 [2].

[tnr,tnrFreq,isProminent] = acousticToneToNoiseRatio(audioIn,fs) also returns a logical vector specifying which tones are prominent according to ECMA-418-1.

[tnr,tnrFreq,isProminent,timestamps] = acousticToneToNoiseRatio(audioIn,fs,TimeVarying=true) detects tones and measures TNR in a time-varying signal. This syntax also returns a vector of timestamps for each time segment in the input signal.

[___] = acousticToneToNoiseRatio(___,Name=Value) specifies options using one or more name-value arguments. For example, acousticToneToNoiseRatio(x,fs,ReturnProminentOnly=true) returns only the prominent tones detected in the input signal.

acousticToneToNoiseRatio(___) with no output arguments plots the detected tones and their TNRs.

Examples

Detect and Measure Tones with Tone-To-Noise Ratio

Read in an audio signal containing the sound from a turbine.

[x,fs] = audioread("Turbine-16-44p1-mono-22secs.wav");

Listen to the first three seconds of the audio signal.

sound(x(1:fs*3),fs)

Use acousticToneToNoiseRatio to detect and measure tones in the stationary signal.

• tnrFreq contains the frequencies of the detected tones in Hz.

• tnr contains the tone-to-noise ratios of the detected tones.

[tnr,tnrFreq] = acousticToneToNoiseRatio(x,fs);

Call acousticToneToNoiseRatio with no output arguments to plot the tones. The plot contains data tips with information about the detected tones.

acousticToneToNoiseRatio(x,fs)

Detect Tones in a Time-Varying Audio Signal

Read in an audio signal containing a train whistle and listen to it.

[x,fs] = audioread("TrainWhistle-16-44p1-mono-9secs.wav");
sound(x,fs)

Call acousticToneToNoiseRatio with TimeVarying set to true to detect the tones in the time-varying signal.

[tnr,tnrFreq,isProminent,timestamps] = acousticToneToNoiseRatio(x,fs,TimeVarying=true);

Call acousticToneToNoiseRatio with no output arguments to plot the time-varying tones and their tone-to-noise ratios. The plot contains data tips with information about the detected tones.

acousticToneToNoiseRatio(x,fs,TimeVarying=true)

Find Prominent Tones with the Tone-To-Noise Ratio

Read in an audio signal containing the sound from a turbine.

[x,fs] = audioread("Turbine-16-44p1-mono-22secs.wav");

Use acousticToneToNoiseRatio to detect the tones in the signal. Specify the third output argument to get a logical vector indicating which tones are prominent.

[tnr,tnrFreq,isProminent] = acousticToneToNoiseRatio(x,fs);

Use isProminent to get the frequencies of the prominent tones. This signal has one prominent tone.

tnrFreq(isProminent)

ans = 
1.0713e+03

Specify Custom Parameters for Measuring Tone-To-Noise Ratio

Read in an audio signal containing the sound from a turbine.

[x,fs] = audioread("Turbine-16-44p1-mono-22secs.wav");

Specify custom parameters for the spectral analysis performed by acousticToneToNoiseRatio.

nfft = 2048;
window = hamming(nfft,"periodic");
overlap = round(0.75*numel(window));
freqRange = [100 16000];

Call acousticToneToNoiseRatio with the custom parameters and see the resulting plot.

acousticToneToNoiseRatio(x,fs,FFTLength=nfft,Window=window, ...
    OverlapLength=overlap,FrequencyRange=freqRange)

Detect and Measure Tones in Magnitude Spectrum

Read in an audio signal containing the sound from a turbine.

[x,fs] = audioread("Turbine-16-44p1-mono-22secs.wav");

Compute the magnitude spectrum of the stationary signal and convert it to dB.

nfft = 16384;
[pxx,f] = pwelch(x,hann(nfft,"periodic"),nfft/2,nfft,fs);
pxxDB = 10*log10(pxx);

Use acousticToneToNoiseRatio to detect and measure tones in the spectrum.

acousticToneToNoiseRatio(pxxDB,f);

Detect and Measure Tones in Spectrogram

Read in an audio signal containing a train whistle.

[x,fs] = audioread("TrainWhistle-16-44p1-mono-9secs.wav");

Compute the spectrogram of the time-varying signal and convert it to dB.

nfft = 16384;
h = hann(nfft,"periodic");
[~,fs,~,psd] = spectrogram(x,h,nfft/2,nfft,fs);
psdDB = 10*log10(psd);

Use acousticToneToNoiseRatio to detect and measure tones in the spectrogram.

acousticToneToNoiseRatio(psdDB,fs,TimeVarying=true)

Input Arguments

audioIn — Audio input

column vector | matrix

Audio input, specified as a column vector or matrix. The audio input can be a time-domain or frequency-domain signal.

• If fs is a scalar, audioIn must be a column vector, which acousticToneToNoiseRatio interprets as a single-channel time-domain signal.

• If fs is a vector and TimeVarying is false or unspecified, audioIn must be a column vector, which acousticToneToNoiseRatio interprets as a magnitude spectrum in dB.

• If fs is a vector and TimeVarying is true, audioIn must be a matrix, which acousticToneToNoiseRatio interprets as a magnitude spectrogram in dB. Each column in the spectrogram is a spectrum corresponding to a time segment.

Data Types: single | double

fs — Sample rate or frequency vector (Hz)

positive scalar | vector

Sample rate or frequency vector in Hz, specified as a positive scalar or vector.

• If fs is a scalar, it is the sample rate of the time-domain input signal audioIn. To cover the frequency range specified in the ECMA-418-1 [2] standard, the sample rate should be at least 32 kHz.

• If fs is a vector, it contains the frequencies corresponding to each row in the frequency-domain input signal audioIn.

Data Types: single | double

Name-Value Arguments

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.

Before R2021a, use commas to separate each name and value, and enclose Name in quotes.

Example: acousticToneToNoiseRatio(x,fs,FrequencyRange=[500 2000])

TimeVarying — Time-varying input signal

false (default) | true

Time-varying input signal, specified as true or false. If TimeVarying is true, acousticToneToNoiseRatio detects and measures tones in each time segment of the time-varying signal.

• If the input is a time-domain signal, the Window and OverlapLength determine the time segments of the time-varying signal.

• If the input is a frequency-domain signal, each column in the magnitude spectrogram corresponds to a time segment.

Data Types: logical

FFTLength — FFT length

min(size(audioIn,1),2^nextpow2(0.2*fs)) (default) | positive integer

FFT length, specified as a positive integer greater than or equal to the length of Window.

This argument does not apply if the input is a frequency-domain signal.

Data Types: single | double

Window — Analysis window

hann(FFTLength,"periodic") (default) | vector

Analysis window, specified as a vector.

This argument does not apply if the input is a frequency-domain signal.

Data Types: single | double

OverlapLength — Overlap length

round(numel(Window)/2) (default) | positive integer

Overlap length, specified as a positive integer less than the length of Window. This argument specifies the number of samples overlap between adjacent windows.

This argument does not apply if the input is a frequency-domain signal.

Data Types: single | double

FrequencyRange — Frequency range (Hz)

[89.1 11200] (default) | two-element row vector

Frequency range in Hz over which to analyze tones, specified as a two-element row vector.

Data Types: single | double

ReturnProminentOnly — Flag to only return prominent tones

false (default) | true

Flag to only return prominent tones, specified as true or false. If you specify this argument as true, acousticToneToNoiseRatio returns only tones that are prominent according to the ECMA-418-1 standard [2].

Data Types: logical

CombineProximalTones — Flag to combine proximal tones

true (default) | false

Flag to combine proximal tones, specified as true or false. If this argument is unspecified or true, acousticToneToNoiseRatio combines tones that are close enough to be perceived as a single tone according to the proximity criteria of the ECMA-418-1 standard.

Data Types: single | double

Output Arguments

tnr — Tone-to-noise ratios

vector | cell array of vectors

Tone-to-noise ratios (TNR) of detected tones, returned as a vector in decreasing order. The TNR is a metric that measures perceived tonality, and it is defined in the ECMA-418-1 standard [2]. Each value in the vector is the TNR of the corresponding frequency in tnrFreq.

If TimeVarying is true, acousticToneToNoiseRatio returns tnr as a cell array of vectors where each cell corresponds to a time segment.

tnrFreq — Frequencies of detected tones (Hz)

vector | cell array of vectors

Frequencies of the detected tones in Hz, returned as a vector where each element corresponds to a TNR value in tnr.

If TimeVarying is true, acousticToneToNoiseRatio returns tnrFreq as a cell array of vectors where each cell corresponds to a time segment.

isProminent — Prominent tones

vector | cell array of vectors

Prominent tones according to the ECMA-418-1 standard, returned as a logical vector with a value corresponding to each tone in tnrFreq indicating whether it is prominent.

If TimeVarying is true, acousticToneToNoiseRatio returns isProminent as a cell array of vectors where each cell corresponds to a time segment.

timestamps — Timestamps

vector

Timestamps of each time segment in the input signal in seconds, returned as a vector.

This output argument applies only if you set TimeVarying to true.

Algorithms

The acousticToneToNoiseRatio function detects the tones in the input signal using the methods proposed in [1].

References

[1] Bray, Wade R. “Methods for Automating Prominent Tone Evaluation and for Considering Variations with Time or Other Reference Quantities.” The Journal of the Acoustical Society of America, vol. 123, no. 5, May 2008, pp. 3685–3685. DOI.org (Crossref), https://doi.org/10.1121/1.2935056.

[2] ECMA-418-1. "Psychoacoustic metrics for ITT equipment — Part 1 (prominent discrete tones)." Ecma International. https://ecma-international.org/publications-and-standards/standards/ecma-418/.

Extended Capabilities

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

Version History

Introduced in R2023b