Table 1. Comparison of proposed SNN and other models on the isolated spoken digit classification task.
RC: Reservoir computing; FC: Fully-connected layer; CSNN: Convolutional spiking neural network; HTM: Hierarchical temporal memory; RNN: Recurrent neural network; DN: Delta network.
| Model | Architecture | Learning method | Dataset | Speakers | Accuracy |
|---|---|---|---|---|---|
| SNNs | |||||
| Verstraeten et al. [19] | RC | Pseudo matrix inversion | TI46 | 5 | >97.5 |
| Zhang et al. [21] | RC | Abstract learning rule | TI46 | 16 | 92.3 |
| Wade et al. [27] | FC | STDP/BCM | TI46 | 16 | 95.25 |
| Tavanaei et al. [28] | FC | Hebbian/anti-Hebbian STDP | Aurora | 50 | 91 |
| Tavanaei et al. [29] | CSNN | STDP | Aurora | >50 | 96 |
| Dibazar et al. [50] | FC | Backpropagation | TIDIGITS | <80 | 85.1 |
| Our model | CSNN | STDP | TIDIGITS | 200 | 97.5 |
| ANNs | |||||
| van Doremalen et al. [51] | HTM | Coincidence memorization | TIDIGITS | 150 | 91.43 |
| Neil et al. [52] | RNN | Backpropagation | TIDIGITS | 200 | 96.1 |
| Neil et al. [53] | DN | Backpropagation | TIDIGITS | 200 | 97.5 |