Speech Enhancement and Assessment Resource (SpEAR)

Home | Demos | Research | Database and Tools | Publications | People | Links
SpEAR description | SpEAR beta release

SpEAR consists of a speech database and toolkit for assessing performance of speech enhancement algorithms.

SpEAR Database: Noisy Speech Recordings

Recorded speech and recorded noise: acoustically combined and re-recorded.
(Technical Details)

Original TIMIT Clean Room Reference F16 Noise Factory Noise Pink Noise White Noise Volvo Noise Bursting Noise
bigtips Inf. dB 2.74 dB 3.33 dB 2.16 dB 2.37 dB 7.22 dB 0.16 dB
peaches Inf. dB 3.26 dB 4.07 dB 2.77 dB 3.32 dB 7.68 dB 0.26 dB
draw Inf. dB 4.35 dB 5.17 dB 3.87 dB 4.08 dB 8.79 dB 1.37 dB
scholars Inf. dB 9.21 dB 8.98 dB 8.55 dB 8.77 dB 13.75 dB 6.47 dB

Original TIMIT bigtips
Clean Room Reference Inf. dB
F16 Noise at SNR: -5.93 dB
-4.62 dB
-3.63 dB
-2.65 dB
-0.67 dB
+0.37 dB
+1.29 dB
+2.29 dB
+3.26 dB
+5.23 dB
+7.20 dB
+9.17 dB
+10.16 dB
+11.15 dB
+12.14 dB
+14.11 dB
+16.08 dB
+17.05 dB
 
Original Sound vega
Clean Room Reference Inf. dB
Pink Noise at SNR: -0.92 dB
+1.05 dB
+3.03 dB
+5.00 dB
+6.97 dB
+8.94 dB
+12.89 dB
+14.85 dB
+16.81 dB
+18.76 dB