The low frame rate algorithm is a simple method for subsampling in speech processing. At time $i$, the acoustic feature vector $f_i$ and the two prior vectors $f_{i-1}$ and $f_{i-2}$ are concatenated. This creates a new vector that is three times longer, after which the original $f_{i-1}$ and $f_{i-2}$ vectors are discarded. This results in a higher-dimensional vector with a shorter overall sequence length, $n = \frac{t}{3}$. For example, instead of processing a 40-dimensional acoustic feature vector every 10 ms, this method yields a 120-dimensional vector every 30 ms.