Keyword Spotting

sherpa-onnx supports keyword spotting through the SherpaOnnxCreateKeywordSpotter() API. It uses streaming transducer models with keyword lists.

See also

SherpaOnnxCreateKeywordSpotter, SherpaOnnxKeywordSpotterConfig

SherpaOnnxKeywordSpotterConfig config;
memset(&config, 0, sizeof(config));
config.model_config.transducer.encoder =
    "./sherpa-onnx-kws-zipformer-wenetspeech-3.3M-2024-01-01-mobile/"
    "encoder-epoch-12-avg-2-chunk-16-left-64.int8.onnx";
config.model_config.transducer.decoder =
    "./sherpa-onnx-kws-zipformer-wenetspeech-3.3M-2024-01-01-mobile/"
    "decoder-epoch-12-avg-2-chunk-16-left-64.onnx";
config.model_config.transducer.joiner =
    "./sherpa-onnx-kws-zipformer-wenetspeech-3.3M-2024-01-01-mobile/"
    "joiner-epoch-12-avg-2-chunk-16-left-64.int8.onnx";
config.model_config.tokens =
    "./sherpa-onnx-kws-zipformer-wenetspeech-3.3M-2024-01-01-mobile/tokens.txt";
config.model_config.provider = "cpu";
config.model_config.num_threads = 1;
config.keywords_file =
    "./sherpa-onnx-kws-zipformer-wenetspeech-3.3M-2024-01-01-mobile/"
    "test_wavs/test_keywords.txt";
config.max_active_paths = 4;
config.keywords_score = 3.0f;
config.keywords_threshold = 0.1f;
 
const SherpaOnnxKeywordSpotter *kws =
    SherpaOnnxCreateKeywordSpotter(&config);

Model package: sherpa-onnx-kws-zipformer-wenetspeech-3.3M-2024-01-01-mobile

Example source: kws-c-api.c