Keyword Spotting
In speech processing, keyword spotting deals with the identification of keywords in utterances.
( Image credit: Simon Grest )
Papers
No papers found.
All datasetsQUESSTGoogle Speech Commandshey SiriFKDGoogle Speech Commands V2 35TensorFlowVoxForgeGoogle Speech Commands (v2)Google Speech Commands V2 12
Benchmark Results
| # | Model | Metric | Claimed | Verified | Status |
|---|---|---|---|---|---|
| 1 | NNI non-filtered(for the development set) | Cnxe | 6.09 | — | Unverified |
| 2 | NNI Choi(for the development set) | Cnxe | 5.89 | — | Unverified |
| 3 | NTU rnn (eval) | Cnxe | 2.01 | — | Unverified |
| 4 | NTU dtw (eval) | Cnxe | 2.01 | — | Unverified |
| 5 | NTU dtw (dev) | Cnxe | 2.01 | — | Unverified |
| 6 | NTU rnn (dev) | Cnxe | 2.01 | — | Unverified |
| 7 | ELiRF SDTW (eval) | Cnxe | 1.19 | — | Unverified |
| 8 | ELiRF SDTW-avg (eval) | Cnxe | 1.07 | — | Unverified |
| 9 | ELiRF SDTW (dev) | Cnxe | 1.07 | — | Unverified |
| 10 | CUNY [Subseq+MFCC] (eval) | Cnxe | 1.07 | — | Unverified |
| # | Model | Metric | Claimed | Verified | Status |
|---|---|---|---|---|---|
| 1 | WaveFormer | Google Speech Commands V2 12 | 98.8 | — | Unverified |
| 2 | QNN | Google Speech Commands V2 35 | 98.6 | — | Unverified |
| 3 | TripletLoss-res15 | Google Speech Commands V1 12 | 98.56 | — | Unverified |
| 4 | M2D | Google Speech Commands V2 35 | 98.5 | — | Unverified |
| 5 | EAT-S | Google Speech Commands V2 35 | 98.15 | — | Unverified |
| 6 | Audio Spectrogram Transformer | Google Speech Commands V2 35 | 98.11 | — | Unverified |
| 7 | EdgeCRNN 2.0× | Google Speech Commands V2 12 | 98.05 | — | Unverified |
| 8 | BC-ResNet-8 | Google Speech Commands V1 12 | 98 | — | Unverified |
| 9 | HTS-AT | Google Speech Commands V2 35 | 98 | — | Unverified |
| 10 | Wav2KWS | Google Speech Commands V1 12 | 97.9 | — | Unverified |
| # | Model | Metric | Claimed | Verified | Status |
|---|---|---|---|---|---|
| 1 | Stacked 1D CNN | Error Rate | 1.99 | — | Unverified |
| 2 | End-to-end DNN-HMM | Error Rate | 1.7 | — | Unverified |
| 3 | HEiMDaL | Error Rate | 0.45 | — | Unverified |
| # | Model | Metric | Claimed | Verified | Status |
|---|---|---|---|---|---|
| 1 | Res26 | Accuracy | 95.88 | — | Unverified |
| 2 | EfficientNet-A0 + SA + TL | Accuracy | 95.83 | — | Unverified |
| # | Model | Metric | Claimed | Verified | Status |
|---|---|---|---|---|---|
| 1 | QuaternionNeuralNetwork | Accuracy (10-fold) | 98.53 | — | Unverified |
| 2 | SSAMBA | Accuracy (10-fold) | 97.4 | — | Unverified |
| # | Model | Metric | Claimed | Verified | Status |
|---|---|---|---|---|---|
| 1 | TensorFlow's model version 2 | TFMA | 89.7 | — | Unverified |
| 2 | TensorFlow's model version 1 | TFMA | 85.4 | — | Unverified |
| # | Model | Metric | Claimed | Verified | Status |
|---|---|---|---|---|---|
| 1 | 2D-ConvNet | Accuracy (%) | 95.4 | — | Unverified |
| 2 | 1D-ConvNet | Accuracy (%) | 93.7 | — | Unverified |
| # | Model | Metric | Claimed | Verified | Status |
|---|---|---|---|---|---|
| 1 | Quaternion Neural Networks | Accuracy(10-fold) | 98.53 | — | Unverified |
| # | Model | Metric | Claimed | Verified | Status |
|---|---|---|---|---|---|
| 1 | MicroNet-KWS-L | Accuracy | 95.3 | — | Unverified |