The signals in this page demonstrate the doppler-to-speech conversion described in our ICASSP 2010 submission.
There are four columns in the following table. The first column is the Doppler signal. The second column is the speech signal that was recorded along with the Doppler. The third is the speech symthesized from the Doppler, using the power and F0 estimates from the speech signal in column 2. The signals in the final column are synthesized entirely from the Doppler. They are whispered, as no F0 was used.
The training data used for this experiment are available from the link below the table.
| Utterance ID | Doppler Signal | Recorded Speech | Synthesis from Doppler, using F0 and energy from the recorded speech | Synthesis from only the Doppler signal |
| SI2046 | sample | sample | sample | sample |
| SI1738 | sample | sample | sample | sample |
| SI2163 | sample | sample | sample | sample |
| SX232 | sample | sample | sample | sample |
| SX323 | sample | sample | sample | sample |
| SI1464 | sample | sample | sample | sample |
| SI1594 | sample | sample | sample | sample |
| SI1295 | sample | sample | sample | sample |
| SI1955 | sample | sample | sample | sample |
| SI923 | sample | sample | sample | sample |
| SI558 | sample | sample | sample | sample |
| SI647 | sample | sample | sample | sample |
| SX109 | sample | sample | sample | sample |
| SX434 | sample | sample | sample | sample |
| SI2340 | sample | sample | sample | sample |
| SI1804 | sample | sample | sample | sample |
| SI837 | sample | sample | sample | sample |
| SI1132 | sample | sample | sample | sample |
The set of training data used for the experiment. Each recording is stereo. One channel contains the heterodyned and down-sampled Doppler ultrasound signal. The other contains the audio.