Table 2 Comparison of ASR and AVSR technologies for dysarthric speech
From: A survey of technologies for automatic Dysarthric speech recognition
Type of Model | ASR or AVSR | Training Time | Required Video? | Parameter Size of Model | Required GPU? |
|---|---|---|---|---|---|
Machine Learning | ASR | Hours Level | No | ≤Kilobyte Level | No Need |
Deep Learning | ASR | Days Level | No | >Mbyte Level | Yes |
Deep Learning | AVSR | Days (or even Weeks Level) | Yes | > > Mbyte Level (Gigabyte Level) | Yes |