Add C API for spoken language identification. (#695)

sherpa-onnx/c-api/c-api.cc

@@ -6,6 +6,7 @@
 
 #include <algorithm>
 #include <memory>
+#include <string>
 #include <utility>
 #include <vector>
 
@@ -16,7 +17,9 @@
 #include "sherpa-onnx/csrc/offline-recognizer.h"
 #include "sherpa-onnx/csrc/offline-tts.h"
 #include "sherpa-onnx/csrc/online-recognizer.h"
+#include "sherpa-onnx/csrc/spoken-language-identification.h"
 #include "sherpa-onnx/csrc/voice-activity-detector.h"
+#include "sherpa-onnx/csrc/wave-reader.h"
 #include "sherpa-onnx/csrc/wave-writer.h"
 
 struct SherpaOnnxOnlineRecognizer {
@@ -859,3 +862,97 @@ int32_t SherpaOnnxWriteWave(const float *samples, int32_t n,
                             int32_t sample_rate, const char *filename) {
   return sherpa_onnx::WriteWave(filename, sample_rate, samples, n);
 }
+
+const SherpaOnnxWave *SherpaOnnxReadWave(const char *filename) {
+  int32_t sample_rate = -1;
+  bool is_ok = false;
+  std::vector<float> samples =
+      sherpa_onnx::ReadWave(filename, &sample_rate, &is_ok);
+  if (!is_ok) {
+    return nullptr;
+  }
+
+  float *c_samples = new float[samples.size()];
+  std::copy(samples.begin(), samples.end(), c_samples);
+
+  SherpaOnnxWave *wave = new SherpaOnnxWave;
+  wave->samples = c_samples;
+  wave->sample_rate = sample_rate;
+  wave->num_samples = samples.size();
+  return wave;
+}
+
+void SherpaOnnxFreeWave(const SherpaOnnxWave *wave) {
+  if (wave) {
+    delete[] wave->samples;
+    delete wave;
+  }
+}
+
+struct SherpaOnnxSpokenLanguageIdentification {
+  std::unique_ptr<sherpa_onnx::SpokenLanguageIdentification> impl;
+};
+
+const SherpaOnnxSpokenLanguageIdentification *
+SherpaOnnxCreateSpokenLanguageIdentification(
+    const SherpaOnnxSpokenLanguageIdentificationConfig *config) {
+  sherpa_onnx::SpokenLanguageIdentificationConfig slid_config;
+  slid_config.whisper.encoder = SHERPA_ONNX_OR(config->whisper.encoder, "");
+  slid_config.whisper.decoder = SHERPA_ONNX_OR(config->whisper.decoder, "");
+  slid_config.whisper.tail_paddings =
+      SHERPA_ONNX_OR(config->whisper.tail_paddings, -1);
+  slid_config.num_threads = SHERPA_ONNX_OR(config->num_threads, 1);
+  slid_config.debug = config->debug;
+  slid_config.provider = SHERPA_ONNX_OR(config->provider, "cpu");
+
+  if (slid_config.debug) {
+    SHERPA_ONNX_LOGE("%s\n", slid_config.ToString().c_str());
+  }
+
+  if (!slid_config.Validate()) {
+    SHERPA_ONNX_LOGE("Errors in config");
+    return nullptr;
+  }
+
+  SherpaOnnxSpokenLanguageIdentification *slid =
+      new SherpaOnnxSpokenLanguageIdentification;
+  slid->impl =
+      std::make_unique<sherpa_onnx::SpokenLanguageIdentification>(slid_config);
+
+  return slid;
+}
+
+void SherpaOnnxDestroySpokenLanguageIdentification(
+    const SherpaOnnxSpokenLanguageIdentification *slid) {
+  delete slid;
+}
+
+SherpaOnnxOfflineStream *
+SherpaOnnxSpokenLanguageIdentificationCreateOfflineStream(
+    const SherpaOnnxSpokenLanguageIdentification *slid) {
+  SherpaOnnxOfflineStream *stream =
+      new SherpaOnnxOfflineStream(slid->impl->CreateStream());
+  return stream;
+}
+
+const SherpaOnnxSpokenLanguageIdentificationResult *
+SherpaOnnxSpokenLanguageIdentificationCompute(
+    const SherpaOnnxSpokenLanguageIdentification *slid,
+    const SherpaOnnxOfflineStream *s) {
+  std::string lang = slid->impl->Compute(s->impl.get());
+  char *c_lang = new char[lang.size() + 1];
+  std::copy(lang.begin(), lang.end(), c_lang);
+  c_lang[lang.size()] = '\0';
+  SherpaOnnxSpokenLanguageIdentificationResult *r =
+      new SherpaOnnxSpokenLanguageIdentificationResult;
+  r->lang = c_lang;
+  return r;
+}
+
+void SherpaOnnxDestroySpokenLanguageIdentificationResult(
+    const SherpaOnnxSpokenLanguageIdentificationResult *r) {
+  if (r) {
+    delete[] r->lang;
+    delete r;
+  }
+}

sherpa-onnx/c-api/c-api.h

@@ -820,6 +820,76 @@ SHERPA_ONNX_API int32_t SherpaOnnxWriteWave(const float *samples, int32_t n,
                                             int32_t sample_rate,
                                             const char *filename);
 
+SHERPA_ONNX_API typedef struct SherpaOnnxWave {
+  // samples normalized to the range [-1, 1]
+  const float *samples;
+  int32_t sample_rate;
+  int32_t num_samples;
+} SherpaOnnxWave;
+
+// Return a NULL pointer on error. It supports only standard WAVE file.
+// Each sample should be 16-bit. It supports only single channel..
+//
+// If the returned pointer is not NULL, the user has to invoke
+// SherpaOnnxFreeWave() to free the returned pointer to avoid memory leak.
+SHERPA_ONNX_API const SherpaOnnxWave *SherpaOnnxReadWave(const char *filename);
+
+SHERPA_ONNX_API void SherpaOnnxFreeWave(const SherpaOnnxWave *wave);
+
+// Spoken language identification
+
+SHERPA_ONNX_API typedef struct
+    SherpaOnnxSpokenLanguageIdentificationWhisperConfig {
+  const char *encoder;
+  const char *decoder;
+  int32_t tail_paddings;
+} SherpaOnnxSpokenLanguageIdentificationWhisperConfig;
+
+SHERPA_ONNX_API typedef struct SherpaOnnxSpokenLanguageIdentificationConfig {
+  SherpaOnnxSpokenLanguageIdentificationWhisperConfig whisper;
+  int32_t num_threads;
+  int32_t debug;
+  const char *provider;
+} SherpaOnnxSpokenLanguageIdentificationConfig;
+
+SHERPA_ONNX_API typedef struct SherpaOnnxSpokenLanguageIdentification
+    SherpaOnnxSpokenLanguageIdentification;
+
+// Create an instance of SpokenLanguageIdentification.
+// The user has to invoke SherpaOnnxDestroySpokenLanguageIdentification()
+// to free the returned pointer to avoid memory leak.
+SHERPA_ONNX_API const SherpaOnnxSpokenLanguageIdentification *
+SherpaOnnxCreateSpokenLanguageIdentification(
+    const SherpaOnnxSpokenLanguageIdentificationConfig *config);
+
+SHERPA_ONNX_API void SherpaOnnxDestroySpokenLanguageIdentification(
+    const SherpaOnnxSpokenLanguageIdentification *slid);
+
+// The user has to invoke DestroyOfflineStream()
+// to free the returned pointer to avoid memory leak
+SHERPA_ONNX_API SherpaOnnxOfflineStream *
+SherpaOnnxSpokenLanguageIdentificationCreateOfflineStream(
+    const SherpaOnnxSpokenLanguageIdentification *slid);
+
+SHERPA_ONNX_API typedef struct SherpaOnnxSpokenLanguageIdentificationResult {
+  // en for English
+  // de for German
+  // zh for Chinese
+  // es for Spanish
+  // ...
+  const char *lang;
+} SherpaOnnxSpokenLanguageIdentificationResult;
+
+// The user has to invoke SherpaOnnxDestroySpokenLanguageIdentificationResult()
+// to free the returned pointer to avoid memory leak
+SHERPA_ONNX_API const SherpaOnnxSpokenLanguageIdentificationResult *
+SherpaOnnxSpokenLanguageIdentificationCompute(
+    const SherpaOnnxSpokenLanguageIdentification *slid,
+    const SherpaOnnxOfflineStream *s);
+
+SHERPA_ONNX_API void SherpaOnnxDestroySpokenLanguageIdentificationResult(
+    const SherpaOnnxSpokenLanguageIdentificationResult *r);
+
 #if defined(__GNUC__)
 #pragma GCC diagnostic pop
 #endif