Table 2.
Statistical comparison between the intelligibility of unprocessed noisy speech and that of processed speech by five noise-reduction algorithms for Japanese. Algorithms indicated with “E” were found to be equally intelligible to noisy speech, algorithms indicated with “L” provided low intelligibility scores, and algorithms indicated with “B” improved intelligibility.
| 0 dB | 5 dB | |||||
|---|---|---|---|---|---|---|
| Algorithm | White | Babble | Car | White | Babble | Car |
| KLT | L | L | L | E | L | E |
| logMMSE | E | L | E | L | L | E |
| logMMSE-SPU | L | L | L | L | L | L |
| MB | L | L | L | L | L | L |
| Wiener-as | B | L | B | B | L | B |