diff --git a/.github/scripts/test-nodejs-addon-npm.sh b/.github/scripts/test-nodejs-addon-npm.sh index 7fde8726..e2c0a95e 100755 --- a/.github/scripts/test-nodejs-addon-npm.sh +++ b/.github/scripts/test-nodejs-addon-npm.sh @@ -10,6 +10,16 @@ arch=$(node -p "require('os').arch()") platform=$(node -p "require('os').platform()") node_version=$(node -p "process.versions.node.split('.')[0]") +echo "----------non-streaming ASR dolphin CTC----------" + +curl -SL -O https://github.com/k2-fsa/sherpa-onnx/releases/download/asr-models/sherpa-onnx-dolphin-base-ctc-multi-lang-int8-2025-04-02.tar.bz2 +tar xvf sherpa-onnx-dolphin-base-ctc-multi-lang-int8-2025-04-02.tar.bz2 +rm sherpa-onnx-dolphin-base-ctc-multi-lang-int8-2025-04-02.tar.bz2 + +node ./test_asr_non_streaming_dolphin_ctc.js + +rm -rf sherpa-onnx-dolphin-base-ctc-multi-lang-int8-2025-04-02 + echo "----------non-streaming speech denoiser----------" curl -SL -O https://github.com/k2-fsa/sherpa-onnx/releases/download/speech-enhancement-models/gtcrn_simple.onnx diff --git a/harmony-os/SherpaOnnxHar/sherpa_onnx/Index.ets b/harmony-os/SherpaOnnxHar/sherpa_onnx/Index.ets index 84286294..7b173404 100644 --- a/harmony-os/SherpaOnnxHar/sherpa_onnx/Index.ets +++ b/harmony-os/SherpaOnnxHar/sherpa_onnx/Index.ets @@ -6,6 +6,7 @@ export { CircularBuffer, SileroVadConfig, SpeechSegment, Vad, VadConfig, } from export { Samples, OfflineStream, FeatureConfig, + OfflineDolphinModelConfig, OfflineTransducerModelConfig, OfflineParaformerModelConfig, OfflineNemoEncDecCtcModelConfig, diff --git a/harmony-os/SherpaOnnxHar/sherpa_onnx/src/main/cpp/non-streaming-asr.cc b/harmony-os/SherpaOnnxHar/sherpa_onnx/src/main/cpp/non-streaming-asr.cc index af14bdc1..c2583c32 100644 --- a/harmony-os/SherpaOnnxHar/sherpa_onnx/src/main/cpp/non-streaming-asr.cc +++ b/harmony-os/SherpaOnnxHar/sherpa_onnx/src/main/cpp/non-streaming-asr.cc @@ -44,6 +44,22 @@ static SherpaOnnxOfflineParaformerModelConfig GetOfflineParaformerModelConfig( return c; } +static SherpaOnnxOfflineDolphinModelConfig GetOfflineDolphinfig( + Napi::Object obj) { + SherpaOnnxOfflineDolphinModelConfig c; + memset(&c, 0, sizeof(c)); + + if (!obj.Has("dolphin") || !obj.Get("dolphin").IsObject()) { + return c; + } + + Napi::Object o = obj.Get("dolphin").As(); + + SHERPA_ONNX_ASSIGN_ATTR_STR(model, model); + + return c; +} + static SherpaOnnxOfflineNemoEncDecCtcModelConfig GetOfflineNeMoCtcModelConfig( Napi::Object obj) { SherpaOnnxOfflineNemoEncDecCtcModelConfig c; @@ -168,6 +184,7 @@ static SherpaOnnxOfflineModelConfig GetOfflineModelConfig(Napi::Object obj) { c.sense_voice = GetOfflineSenseVoiceModelConfig(o); c.moonshine = GetOfflineMoonshineModelConfig(o); c.fire_red_asr = GetOfflineFireRedAsrModelConfig(o); + c.dolphin = GetOfflineDolphinfig(o); SHERPA_ONNX_ASSIGN_ATTR_STR(tokens, tokens); SHERPA_ONNX_ASSIGN_ATTR_INT32(num_threads, numThreads); @@ -292,6 +309,8 @@ CreateOfflineRecognizerWrapper(const Napi::CallbackInfo &info) { SHERPA_ONNX_DELETE_C_STR(c.model_config.fire_red_asr.encoder); SHERPA_ONNX_DELETE_C_STR(c.model_config.fire_red_asr.decoder); + SHERPA_ONNX_DELETE_C_STR(c.model_config.dolphin.model); + SHERPA_ONNX_DELETE_C_STR(c.model_config.tokens); SHERPA_ONNX_DELETE_C_STR(c.model_config.provider); SHERPA_ONNX_DELETE_C_STR(c.model_config.model_type); diff --git a/harmony-os/SherpaOnnxHar/sherpa_onnx/src/main/ets/components/NonStreamingAsr.ets b/harmony-os/SherpaOnnxHar/sherpa_onnx/src/main/ets/components/NonStreamingAsr.ets index d3f849cc..94437ac3 100644 --- a/harmony-os/SherpaOnnxHar/sherpa_onnx/src/main/ets/components/NonStreamingAsr.ets +++ b/harmony-os/SherpaOnnxHar/sherpa_onnx/src/main/ets/components/NonStreamingAsr.ets @@ -45,6 +45,10 @@ export class OfflineNemoEncDecCtcModelConfig { public model: string = ''; } +export class OfflineDolphinModelConfig { + public model: string = ''; +} + export class OfflineWhisperModelConfig { public encoder: string = ''; public decoder: string = ''; @@ -86,6 +90,7 @@ export class OfflineModelConfig { public telespeechCtc: string = ''; public senseVoice: OfflineSenseVoiceModelConfig = new OfflineSenseVoiceModelConfig(); public moonshine: OfflineMoonshineModelConfig = new OfflineMoonshineModelConfig(); + public dolphin: OfflineDolphinModelConfig = new OfflineDolphinModelConfig(); } export class OfflineLMConfig { @@ -159,4 +164,4 @@ export class OfflineRecognizer { return r; } -} \ No newline at end of file +} diff --git a/nodejs-addon-examples/README.md b/nodejs-addon-examples/README.md index 1fecaa04..09ec0421 100644 --- a/nodejs-addon-examples/README.md +++ b/nodejs-addon-examples/README.md @@ -122,6 +122,7 @@ The following tables list the examples in this folder. |[./test_asr_non_streaming_moonshine.js](./test_asr_non_streaming_moonshine.js)|Non-streaming speech recognition from a file using [Moonshine](https://github.com/usefulsensors/moonshine)| |[./test_vad_with_non_streaming_asr_moonshine.js](./test_vad_with_non_streaming_asr_moonshine.js)| Non-streaming speech recognition from a file using [Moonshine](https://github.com/usefulsensors/moonshine) + [Silero VAD](https://github.com/snakers4/silero-vad)| |[./test_asr_non_streaming_nemo_ctc.js](./test_asr_non_streaming_nemo_ctc.js)|Non-streaming speech recognition from a file using a [NeMo](https://github.com/NVIDIA/NeMo) CTC model with greedy search| +|[./test_asr_non_streaming_dolphin_ctc.js](./test_asr_non_streaming_dolphin_ctc.js)|Non-streaming speech recognition from a file using a [Dolphinhttps://github.com/DataoceanAI/Dolphin]) CTC model with greedy search| |[./test_asr_non_streaming_paraformer.js](./test_asr_non_streaming_paraformer.js)|Non-streaming speech recognition from a file using [Paraformer](https://github.com/alibaba-damo-academy/FunASR)| |[./test_asr_non_streaming_sense_voice.js](./test_asr_non_streaming_sense_voice.js)|Non-streaming speech recognition from a file using [SenseVoice](https://github.com/FunAudioLLM/SenseVoice)| @@ -332,6 +333,16 @@ wget https://github.com/k2-fsa/sherpa-onnx/releases/download/asr-models/silero_v node ./test_vad_with_non_streaming_asr_whisper.js ``` +### Non-streaming speech recognition with Dolphin CTC models + +```bash +wget https://github.com/k2-fsa/sherpa-onnx/releases/download/asr-models/sherpa-onnx-dolphin-base-ctc-multi-lang-int8-2025-04-02.tar.bz2 +tar xvf sherpa-onnx-dolphin-base-ctc-multi-lang-int8-2025-04-02.tar.bz2 +rm sherpa-onnx-dolphin-base-ctc-multi-lang-int8-2025-04-02.tar.bz2 + +node ./test_asr_non_streaming_dolphin_ctc.js +``` + ### Non-streaming speech recognition with NeMo CTC models ```bash diff --git a/nodejs-addon-examples/test_asr_non_streaming_dolphin_ctc.js b/nodejs-addon-examples/test_asr_non_streaming_dolphin_ctc.js new file mode 100644 index 00000000..42f25511 --- /dev/null +++ b/nodejs-addon-examples/test_asr_non_streaming_dolphin_ctc.js @@ -0,0 +1,48 @@ +// Copyright (c) 2025 Xiaomi Corporation +const sherpa_onnx = require('sherpa-onnx-node'); + +// Please download test files from +// https://github.com/k2-fsa/sherpa-onnx/releases/tag/asr-models +const config = { + 'featConfig': { + 'sampleRate': 16000, + 'featureDim': 80, + }, + 'modelConfig': { + 'dolphin': { + 'model': + './sherpa-onnx-dolphin-base-ctc-multi-lang-int8-2025-04-02/model.int8.onnx', + }, + 'tokens': + './sherpa-onnx-dolphin-base-ctc-multi-lang-int8-2025-04-02/tokens.txt', + 'numThreads': 2, + 'provider': 'cpu', + 'debug': 1, + } +}; + +const waveFilename = + './sherpa-onnx-dolphin-base-ctc-multi-lang-int8-2025-04-02/test_wavs/0.wav'; + +const recognizer = new sherpa_onnx.OfflineRecognizer(config); +console.log('Started') +let start = Date.now(); +const stream = recognizer.createStream(); +const wave = sherpa_onnx.readWave(waveFilename); +stream.acceptWaveform({sampleRate: wave.sampleRate, samples: wave.samples}); + +recognizer.decode(stream); +result = recognizer.getResult(stream) +let stop = Date.now(); +console.log('Done') + +const elapsed_seconds = (stop - start) / 1000; +const duration = wave.samples.length / wave.sampleRate; +const real_time_factor = elapsed_seconds / duration; +console.log('Wave duration', duration.toFixed(3), 'seconds') +console.log('Elapsed', elapsed_seconds.toFixed(3), 'seconds') +console.log( + `RTF = ${elapsed_seconds.toFixed(3)}/${duration.toFixed(3)} =`, + real_time_factor.toFixed(3)) +console.log(waveFilename) +console.log('result\n', result)