NOIZEUS: A noisy speech corpus for evaluation of speech enhancement algorithms



A noisy speech corpus (NOIZEUS) was developed to facilitate comparison of speech enhancement algorithms among research groups. The noisy database contains 30 IEEE sentences (produced by three male and three female speakers) corrupted by eight different real-world noises at different SNRs. The noise was taken from the AURORA database and includes suburban train noise, babble, car, exhibition hall, restaurant, street, airport and train-station noise. This corpus is available to researchers free of charge.

The description of this corpus was published in the following paper, which we ask that you cite when using NOIZEUS:

Hu, Y. and Loizou, P. (2007). “Subjective evaluation and comparison of speech enhancement algorithms,” Speech Communication, 49, 588-601. [pdf]

The NOIZEUS corpus was also used by our lab to evaluate the correlations of common objective measures used in speech enhancement. This work  was reported in the following papers:

Hu, Y. and Loizou, P. (2008). “Evaluation of objective quality measures for speech enhancement,IEEE Transactions on Speech and Audio Processing, 16(1), 229-238.  [Matlab code]

Ma, J.,  Hu, Y. and  Loizou, P. (2009). "Objective measures for predicting speech intelligibility in noisy conditions based on new band-importance functions", Journal of the Acoustical Society of America, 125(5), 3387-3405 [ pdf ]

 


Speech Material

 

Thirty sentences from the IEEE sentence database [1] were recorded in a sound-proof booth using Tucker Davis Technologies (TDT) recording equipment. The sentences were produced by three male and three female speakers. The IEEE database (720 sentences) was used as it contains phonetically-balanced sentences with relatively low word-context predictability. The thirty sentences  were selected from the IEEE database so as to include all phonemes in the American English language. The list of sentences recorded for NOIZEUS are given in Table 1. The sentences were originally sampled at 25 kHz and downsampled to 8 kHz.

 

Filtering

 

 To simulate the receiving frequency characteristics of telephone handsets, the speech and noise signals were filtered by the modified Intermediate Reference System (IRS) filters used in ITU-T P.862 [1] for evaluation of the PESQ measure. The filter frequency response is shown in Figure 1.

Figure 1: Frequency response of IRS filter.

 

Adding Noise

 

 Noise is artificially added to the speech signal as follows. The IRS filter is independently applied to the clean and noise signals. The active speech level of the filtered clean speech signal is first determined using the method B of ITU-T P.56 [3].  A noise segment of the same length as the speech signal is randomly cut out of the noise recordings, appropriately scaled to reach the desired SNR level and finally added to the filtered clean speech signal.

 

 

Noise signals were taken from the AURORA database [4] and included the following recordings from different places:

  • Babble (crowd of people)
  • Car
  • Exhibition hall
  • Restaurant
  • Street
  • Airport
  • Train station
  • Train

 

The long-term spectra of the above noises are given in [4]. The noise signals were added to the speech signals at SNRs of 0dB, 5dB, 10dB, and 15dB.

 

Download files

 

 All files were saved in Windows wav format (16 bit PCM, mono).

 

 

Train noise

SNR (dB)

0

5

10

15

Noisy files (.zip)

train_0dB

train_5dB

train_10dB

train_15dB

 

Babble noise

SNR (dB)

0

5

10

15

Noisy files (.zip)

babble_0dB

babble_5dB

babble_10dB

babble_15dB

 

Car noise

SNR (dB)

0

5

10

15

Noisy files (.zip)

car_0dB

car_5dB

car_10dB

car_15dB

 

Exhibition hall noise

SNR (dB)

0

5

10

15

Noisy files (.zip)

exhibition_0dB

exhibition_5dB

exhibition_10dB

exhibition_15dB

 

Restaurant noise

SNR (dB)

0

5

10

15

Noisy files (.zip)

restaurant_0dB

restaurant_5dB

restaurant_10dB

restaurant_15dB

 

Street noise

SNR (dB)

0

5

10

15

Noisy files (.zip)

street_0dB

street_5dB

street_10dB

street_15dB

 

Airport noise

SNR (dB)

0

5

10

15

Noisy files (.zip)

airport_0dB

airport_5dB

airport_10dB

airport_15dB

 

Train station

SNR (dB)

0

5

10

15

Noisy files (.zip)

station_0dB

station_5dB

station_10dB

station_15dB

 

 

References

 

[1]     IEEE Subcommittee (1969). IEEE Recommended Practice for Speech Quality Measurements. IEEE Trans. Audio and Electroacoustics, AU-17(3), 225-246.

[2]     ITU P.862  (2000). Perceptual evaluation of speech quality (PESQ), and objective method for end-to-end speech quality assessment of narrowband telephone networks and speech codecs. ITU-T Recommendation P. 862

[3]    ITU-T P.56 (1993). Objective measurement of active speech level.

 

[4]   H. Hirsch, and D. Pearce (2000). “The Aurora Experimental Framework for the Performance Evaluation of Speech Recognition Systems under Noisy Conditions.” ISCA ITRW ASR2000, Paris, France, September 18-20.

Questions or Feedback about NOIZEUS

Any inquiries or suggestions about NOIZEUS, please direct them to  Dr. Philip Loizou (email: [email protected])


Home