Add speaker identification APIs for HarmonyOS (#1607)

* Add speaker embedding extractor API for HarmonyOS * Add ArkTS API for speaker identification
2024-12-09 19:23:18 +08:00
parent a743a4400f
commit 314545f938
19 changed files with 374 additions and 60 deletions
--- a/sherpa-onnx/c-api/c-api.cc
+++ b/sherpa-onnx/c-api/c-api.cc
@@ -1328,8 +1328,8 @@ struct SherpaOnnxSpeakerEmbeddingExtractor {
  std::unique_ptr<sherpa_onnx::SpeakerEmbeddingExtractor> impl;
 };

-const SherpaOnnxSpeakerEmbeddingExtractor *
-SherpaOnnxCreateSpeakerEmbeddingExtractor(
+static sherpa_onnx::SpeakerEmbeddingExtractorConfig
+GetSpeakerEmbeddingExtractorConfig(
    const SherpaOnnxSpeakerEmbeddingExtractorConfig *config) {
  sherpa_onnx::SpeakerEmbeddingExtractorConfig c;
  c.model = SHERPA_ONNX_OR(config->model, "");
@@ -1342,9 +1342,21 @@ SherpaOnnxCreateSpeakerEmbeddingExtractor(
  }

  if (config->debug) {
+#if __OHOS__
+    SHERPA_ONNX_LOGE("%{public}s\n", c.ToString().c_str());
+#else
    SHERPA_ONNX_LOGE("%s\n", c.ToString().c_str());
+#endif
  }

+  return c;
+}
+
+const SherpaOnnxSpeakerEmbeddingExtractor *
+SherpaOnnxCreateSpeakerEmbeddingExtractor(
+    const SherpaOnnxSpeakerEmbeddingExtractorConfig *config) {
+  auto c = GetSpeakerEmbeddingExtractorConfig(config);
+
  if (!c.Validate()) {
    SHERPA_ONNX_LOGE("Errors in config!");
    return nullptr;
@@ -1983,6 +1995,23 @@ SherpaOnnxVoiceActivityDetector *SherpaOnnxCreateVoiceActivityDetectorOHOS(
  return p;
 }

+const SherpaOnnxSpeakerEmbeddingExtractor *
+SherpaOnnxCreateSpeakerEmbeddingExtractorOHOS(
+    const SherpaOnnxSpeakerEmbeddingExtractorConfig *config,
+    NativeResourceManager *mgr) {
+  if (!mgr) {
+    return SherpaOnnxCreateSpeakerEmbeddingExtractor(config);
+  }
+
+  auto c = GetSpeakerEmbeddingExtractorConfig(config);
+
+  auto p = new SherpaOnnxSpeakerEmbeddingExtractor;
+
+  p->impl = std::make_unique<sherpa_onnx::SpeakerEmbeddingExtractor>(mgr, c);
+
+  return p;
+}
+
 #if SHERPA_ONNX_ENABLE_TTS == 1
 SherpaOnnxOfflineTts *SherpaOnnxCreateOfflineTtsOHOS(
    const SherpaOnnxOfflineTtsConfig *config, NativeResourceManager *mgr) {
--- a/sherpa-onnx/c-api/c-api.h
+++ b/sherpa-onnx/c-api/c-api.h
@@ -1572,6 +1572,11 @@ SherpaOnnxCreateVoiceActivityDetectorOHOS(

 SHERPA_ONNX_API SherpaOnnxOfflineTts *SherpaOnnxCreateOfflineTtsOHOS(
    const SherpaOnnxOfflineTtsConfig *config, NativeResourceManager *mgr);
+
+SHERPA_ONNX_API const SherpaOnnxSpeakerEmbeddingExtractor *
+SherpaOnnxCreateSpeakerEmbeddingExtractorOHOS(
+    const SherpaOnnxSpeakerEmbeddingExtractorConfig *config,
+    NativeResourceManager *mgr);
 #endif

 #if defined(__GNUC__)
--- a/sherpa-onnx/csrc/offline-tts-vits-impl.h
+++ b/sherpa-onnx/csrc/offline-tts-vits-impl.h
@@ -62,9 +62,9 @@ class OfflineTtsVitsImpl : public OfflineTtsImpl {
      for (const auto &f : files) {
        if (config.model.debug) {
 #if __OHOS__
-          SHERPA_ONNX_LOGE("rule far: %s", f.c_str());
-#else
          SHERPA_ONNX_LOGE("rule far: %{public}s", f.c_str());
+#else
+          SHERPA_ONNX_LOGE("rule far: %s", f.c_str());
 #endif
        }
        std::unique_ptr<fst::FarReader<fst::StdArc>> reader(
--- a/sherpa-onnx/csrc/speaker-embedding-extractor-general-impl.h
+++ b/sherpa-onnx/csrc/speaker-embedding-extractor-general-impl.h
@@ -22,11 +22,10 @@ class SpeakerEmbeddingExtractorGeneralImpl
      const SpeakerEmbeddingExtractorConfig &config)
      : model_(config) {}

-#if __ANDROID_API__ >= 9
+  template <typename Manager>
  SpeakerEmbeddingExtractorGeneralImpl(
-      AAssetManager *mgr, const SpeakerEmbeddingExtractorConfig &config)
+      Manager *mgr, const SpeakerEmbeddingExtractorConfig &config)
      : model_(mgr, config) {}
-#endif

  int32_t Dim() const override { return model_.GetMetaData().output_dim; }

@@ -46,9 +45,15 @@ class SpeakerEmbeddingExtractorGeneralImpl
  std::vector<float> Compute(OnlineStream *s) const override {
    int32_t num_frames = s->NumFramesReady() - s->GetNumProcessedFrames();
    if (num_frames <= 0) {
+#if __OHOS__
+      SHERPA_ONNX_LOGE(
+          "Please make sure IsReady(s) returns true. num_frames: %{public}d",
+          num_frames);
+#else
      SHERPA_ONNX_LOGE(
          "Please make sure IsReady(s) returns true. num_frames: %d",
          num_frames);
+#endif
      return {};
    }

@@ -64,8 +69,13 @@ class SpeakerEmbeddingExtractorGeneralImpl
      if (meta_data.feature_normalize_type == "global-mean") {
        SubtractGlobalMean(features.data(), num_frames, feat_dim);
      } else {
+#if __OHOS__
+        SHERPA_ONNX_LOGE("Unsupported feature_normalize_type: %{public}s",
+                         meta_data.feature_normalize_type.c_str());
+#else
        SHERPA_ONNX_LOGE("Unsupported feature_normalize_type: %s",
                         meta_data.feature_normalize_type.c_str());
+#endif
        exit(-1);
      }
    }
--- a/sherpa-onnx/csrc/speaker-embedding-extractor-impl.cc
+++ b/sherpa-onnx/csrc/speaker-embedding-extractor-impl.cc
@@ -3,6 +3,15 @@
 // Copyright (c)  2024  Xiaomi Corporation
 #include "sherpa-onnx/csrc/speaker-embedding-extractor-impl.h"

+#if __ANDROID_API__ >= 9
+#include "android/asset_manager.h"
+#include "android/asset_manager_jni.h"
+#endif
+
+#if __OHOS__
+#include "rawfile/raw_file_manager.h"
+#endif
+
 #include "sherpa-onnx/csrc/macros.h"
 #include "sherpa-onnx/csrc/onnx-utils.h"
 #include "sherpa-onnx/csrc/speaker-embedding-extractor-general-impl.h"
@@ -35,7 +44,11 @@ static ModelType GetModelType(char *model_data, size_t model_data_length,
  if (debug) {
    std::ostringstream os;
    PrintModelMetadata(os, meta_data);
+#if __OHOS__
+    SHERPA_ONNX_LOGE("%{public}s", os.str().c_str());
+#else
    SHERPA_ONNX_LOGE("%s", os.str().c_str());
+#endif
  }

  Ort::AllocatorWithDefaultOptions allocator;
@@ -59,7 +72,11 @@ static ModelType GetModelType(char *model_data, size_t model_data_length,
  } else if (model_type == "nemo") {
    return ModelType::kNeMo;
  } else {
+#if __OHOS__
+    SHERPA_ONNX_LOGE("Unsupported model_type: %{public}s", model_type.c_str());
+#else
    SHERPA_ONNX_LOGE("Unsupported model_type: %s", model_type.c_str());
+#endif
    return ModelType::kUnknown;
  }
 }
@@ -91,10 +108,10 @@ SpeakerEmbeddingExtractorImpl::Create(
  return nullptr;
 }

-#if __ANDROID_API__ >= 9
+template <typename Manager>
 std::unique_ptr<SpeakerEmbeddingExtractorImpl>
 SpeakerEmbeddingExtractorImpl::Create(
-    AAssetManager *mgr, const SpeakerEmbeddingExtractorConfig &config) {
+    Manager *mgr, const SpeakerEmbeddingExtractorConfig &config) {
  ModelType model_type = ModelType::kUnknown;

  {
@@ -120,6 +137,17 @@ SpeakerEmbeddingExtractorImpl::Create(
  // unreachable code
  return nullptr;
 }
+
+#if __ANDROID_API__ >= 9
+template std::unique_ptr<SpeakerEmbeddingExtractorImpl>
+SpeakerEmbeddingExtractorImpl::Create(
+    AAssetManager *mgr, const SpeakerEmbeddingExtractorConfig &config);
+#endif
+
+#if __OHOS__
+template std::unique_ptr<SpeakerEmbeddingExtractorImpl>
+SpeakerEmbeddingExtractorImpl::Create(
+    NativeResourceManager *mgr, const SpeakerEmbeddingExtractorConfig &config);
 #endif

 }  // namespace sherpa_onnx
--- a/sherpa-onnx/csrc/speaker-embedding-extractor-impl.h
+++ b/sherpa-onnx/csrc/speaker-embedding-extractor-impl.h
@@ -9,11 +9,6 @@
 #include <string>
 #include <vector>

-#if __ANDROID_API__ >= 9
-#include "android/asset_manager.h"
-#include "android/asset_manager_jni.h"
-#endif
-
 #include "sherpa-onnx/csrc/speaker-embedding-extractor.h"

 namespace sherpa_onnx {
@@ -25,10 +20,9 @@ class SpeakerEmbeddingExtractorImpl {
  static std::unique_ptr<SpeakerEmbeddingExtractorImpl> Create(
      const SpeakerEmbeddingExtractorConfig &config);

-#if __ANDROID_API__ >= 9
+  template <typename Manager>
  static std::unique_ptr<SpeakerEmbeddingExtractorImpl> Create(
-      AAssetManager *mgr, const SpeakerEmbeddingExtractorConfig &config);
-#endif
+      Manager *mgr, const SpeakerEmbeddingExtractorConfig &config);

  virtual int32_t Dim() const = 0;

--- a/sherpa-onnx/csrc/speaker-embedding-extractor-model.cc
+++ b/sherpa-onnx/csrc/speaker-embedding-extractor-model.cc
@@ -8,6 +8,15 @@
 #include <utility>
 #include <vector>

+#if __ANDROID_API__ >= 9
+#include "android/asset_manager.h"
+#include "android/asset_manager_jni.h"
+#endif
+
+#if __OHOS__
+#include "rawfile/raw_file_manager.h"
+#endif
+
 #include "sherpa-onnx/csrc/macros.h"
 #include "sherpa-onnx/csrc/onnx-utils.h"
 #include "sherpa-onnx/csrc/session.h"
@@ -28,8 +37,8 @@ class SpeakerEmbeddingExtractorModel::Impl {
    }
  }

-#if __ANDROID_API__ >= 9
-  Impl(AAssetManager *mgr, const SpeakerEmbeddingExtractorConfig &config)
+  template <typename Manager>
+  Impl(Manager *mgr, const SpeakerEmbeddingExtractorConfig &config)
      : config_(config),
        env_(ORT_LOGGING_LEVEL_ERROR),
        sess_opts_(GetSessionOptions(config)),
@@ -39,7 +48,6 @@ class SpeakerEmbeddingExtractorModel::Impl {
      Init(buf.data(), buf.size());
    }
  }
-#endif

  Ort::Value Compute(Ort::Value x) const {
    std::array<Ort::Value, 1> inputs = {std::move(x)};
@@ -68,7 +76,11 @@ class SpeakerEmbeddingExtractorModel::Impl {
    if (config_.debug) {
      std::ostringstream os;
      PrintModelMetadata(os, meta_data);
+#if __OHOS__
+      SHERPA_ONNX_LOGE("%{public}s", os.str().c_str());
+#else
      SHERPA_ONNX_LOGE("%s", os.str().c_str());
+#endif
    }

    Ort::AllocatorWithDefaultOptions allocator;  // used in the macro below
@@ -84,8 +96,14 @@ class SpeakerEmbeddingExtractorModel::Impl {
    std::string framework;
    SHERPA_ONNX_READ_META_DATA_STR(framework, "framework");
    if (framework != "wespeaker" && framework != "3d-speaker") {
+#if __OHOS__
+      SHERPA_ONNX_LOGE(
+          "Expect a wespeaker or a 3d-speaker model, given: %{public}s",
+          framework.c_str());
+#else
      SHERPA_ONNX_LOGE("Expect a wespeaker or a 3d-speaker model, given: %s",
                       framework.c_str());
+#endif
      exit(-1);
    }
  }
@@ -111,11 +129,10 @@ SpeakerEmbeddingExtractorModel::SpeakerEmbeddingExtractorModel(
    const SpeakerEmbeddingExtractorConfig &config)
    : impl_(std::make_unique<Impl>(config)) {}

-#if __ANDROID_API__ >= 9
+template <typename Manager>
 SpeakerEmbeddingExtractorModel::SpeakerEmbeddingExtractorModel(
-    AAssetManager *mgr, const SpeakerEmbeddingExtractorConfig &config)
+    Manager *mgr, const SpeakerEmbeddingExtractorConfig &config)
    : impl_(std::make_unique<Impl>(mgr, config)) {}
-#endif

 SpeakerEmbeddingExtractorModel::~SpeakerEmbeddingExtractorModel() = default;

@@ -128,4 +145,14 @@ Ort::Value SpeakerEmbeddingExtractorModel::Compute(Ort::Value x) const {
  return impl_->Compute(std::move(x));
 }

+#if __ANDROID_API__ >= 9
+template SpeakerEmbeddingExtractorModel::SpeakerEmbeddingExtractorModel(
+    AAssetManager *mgr, const SpeakerEmbeddingExtractorConfig &config);
+#endif
+
+#if __OHOS__
+template SpeakerEmbeddingExtractorModel::SpeakerEmbeddingExtractorModel(
+    NativeResourceManager *mgr, const SpeakerEmbeddingExtractorConfig &config);
+#endif
+
 }  // namespace sherpa_onnx
--- a/sherpa-onnx/csrc/speaker-embedding-extractor-model.h
+++ b/sherpa-onnx/csrc/speaker-embedding-extractor-model.h
@@ -6,11 +6,6 @@

 #include <memory>

-#if __ANDROID_API__ >= 9
-#include "android/asset_manager.h"
-#include "android/asset_manager_jni.h"
-#endif
-
 #include "onnxruntime_cxx_api.h"  // NOLINT
 #include "sherpa-onnx/csrc/speaker-embedding-extractor-model-meta-data.h"
 #include "sherpa-onnx/csrc/speaker-embedding-extractor.h"
@@ -22,10 +17,9 @@ class SpeakerEmbeddingExtractorModel {
  explicit SpeakerEmbeddingExtractorModel(
      const SpeakerEmbeddingExtractorConfig &config);

-#if __ANDROID_API__ >= 9
-  SpeakerEmbeddingExtractorModel(AAssetManager *mgr,
+  template <typename Manager>
+  SpeakerEmbeddingExtractorModel(Manager *mgr,
                                 const SpeakerEmbeddingExtractorConfig &config);
-#endif

  ~SpeakerEmbeddingExtractorModel();

--- a/sherpa-onnx/csrc/speaker-embedding-extractor-nemo-impl.h
+++ b/sherpa-onnx/csrc/speaker-embedding-extractor-nemo-impl.h
@@ -22,11 +22,10 @@ class SpeakerEmbeddingExtractorNeMoImpl : public SpeakerEmbeddingExtractorImpl {
      const SpeakerEmbeddingExtractorConfig &config)
      : model_(config) {}

-#if __ANDROID_API__ >= 9
+  template <typename Manager>
  SpeakerEmbeddingExtractorNeMoImpl(
-      AAssetManager *mgr, const SpeakerEmbeddingExtractorConfig &config)
+      Manager *mgr, const SpeakerEmbeddingExtractorConfig &config)
      : model_(mgr, config) {}
-#endif

  int32_t Dim() const override { return model_.GetMetaData().output_dim; }

@@ -54,9 +53,15 @@ class SpeakerEmbeddingExtractorNeMoImpl : public SpeakerEmbeddingExtractorImpl {
  std::vector<float> Compute(OnlineStream *s) const override {
    int32_t num_frames = s->NumFramesReady() - s->GetNumProcessedFrames();
    if (num_frames <= 0) {
+#if __OHOS__
+      SHERPA_ONNX_LOGE(
+          "Please make sure IsReady(s) returns true. num_frames: %{public}d",
+          num_frames);
+#else
      SHERPA_ONNX_LOGE(
          "Please make sure IsReady(s) returns true. num_frames: %d",
          num_frames);
+#endif
      return {};
    }

@@ -72,8 +77,14 @@ class SpeakerEmbeddingExtractorNeMoImpl : public SpeakerEmbeddingExtractorImpl {
      if (meta_data.feature_normalize_type == "per_feature") {
        NormalizePerFeature(features.data(), num_frames, feat_dim);
      } else {
+#if __OHOS__
+        SHERPA_ONNX_LOGE("Unsupported feature_normalize_type: %{public}s",
+                         meta_data.feature_normalize_type.c_str());
+#else
+
        SHERPA_ONNX_LOGE("Unsupported feature_normalize_type: %s",
                         meta_data.feature_normalize_type.c_str());
+#endif
        exit(-1);
      }
    }
--- a/sherpa-onnx/csrc/speaker-embedding-extractor-nemo-model.cc
+++ b/sherpa-onnx/csrc/speaker-embedding-extractor-nemo-model.cc
@@ -8,6 +8,15 @@
 #include <utility>
 #include <vector>

+#if __ANDROID_API__ >= 9
+#include "android/asset_manager.h"
+#include "android/asset_manager_jni.h"
+#endif
+
+#if __OHOS__
+#include "rawfile/raw_file_manager.h"
+#endif
+
 #include "sherpa-onnx/csrc/macros.h"
 #include "sherpa-onnx/csrc/onnx-utils.h"
 #include "sherpa-onnx/csrc/session.h"
@@ -28,8 +37,8 @@ class SpeakerEmbeddingExtractorNeMoModel::Impl {
    }
  }

-#if __ANDROID_API__ >= 9
-  Impl(AAssetManager *mgr, const SpeakerEmbeddingExtractorConfig &config)
+  template <typename Manager>
+  Impl(Manager *mgr, const SpeakerEmbeddingExtractorConfig &config)
      : config_(config),
        env_(ORT_LOGGING_LEVEL_ERROR),
        sess_opts_(GetSessionOptions(config)),
@@ -39,7 +48,6 @@ class SpeakerEmbeddingExtractorNeMoModel::Impl {
      Init(buf.data(), buf.size());
    }
  }
-#endif

  Ort::Value Compute(Ort::Value x, Ort::Value x_lens) const {
    std::array<Ort::Value, 2> inputs = {std::move(x), std::move(x_lens)};
@@ -73,7 +81,11 @@ class SpeakerEmbeddingExtractorNeMoModel::Impl {
    if (config_.debug) {
      std::ostringstream os;
      PrintModelMetadata(os, meta_data);
+#if __OHOS__
+      SHERPA_ONNX_LOGE("%{public}s", os.str().c_str());
+#else
      SHERPA_ONNX_LOGE("%s", os.str().c_str());
+#endif
    }

    Ort::AllocatorWithDefaultOptions allocator;  // used in the macro below
@@ -93,7 +105,12 @@ class SpeakerEmbeddingExtractorNeMoModel::Impl {
    std::string framework;
    SHERPA_ONNX_READ_META_DATA_STR(framework, "framework");
    if (framework != "nemo") {
+#if __OHOS__
+      SHERPA_ONNX_LOGE("Expect a NeMo model, given: %{public}s",
+                       framework.c_str());
+#else
      SHERPA_ONNX_LOGE("Expect a NeMo model, given: %s", framework.c_str());
+#endif
      exit(-1);
    }
  }
@@ -119,11 +136,10 @@ SpeakerEmbeddingExtractorNeMoModel::SpeakerEmbeddingExtractorNeMoModel(
    const SpeakerEmbeddingExtractorConfig &config)
    : impl_(std::make_unique<Impl>(config)) {}

-#if __ANDROID_API__ >= 9
+template <typename Manager>
 SpeakerEmbeddingExtractorNeMoModel::SpeakerEmbeddingExtractorNeMoModel(
-    AAssetManager *mgr, const SpeakerEmbeddingExtractorConfig &config)
+    Manager *mgr, const SpeakerEmbeddingExtractorConfig &config)
    : impl_(std::make_unique<Impl>(mgr, config)) {}
-#endif

 SpeakerEmbeddingExtractorNeMoModel::~SpeakerEmbeddingExtractorNeMoModel() =
    default;
@@ -142,4 +158,14 @@ OrtAllocator *SpeakerEmbeddingExtractorNeMoModel::Allocator() const {
  return impl_->Allocator();
 }

+#if __ANDROID_API__ >= 9
+template SpeakerEmbeddingExtractorNeMoModel::SpeakerEmbeddingExtractorNeMoModel(
+    AAssetManager *mgr, const SpeakerEmbeddingExtractorConfig &config);
+#endif
+
+#if __OHOS__
+template SpeakerEmbeddingExtractorNeMoModel::SpeakerEmbeddingExtractorNeMoModel(
+    NativeResourceManager *mgr, const SpeakerEmbeddingExtractorConfig &config);
+#endif
+
 }  // namespace sherpa_onnx
--- a/sherpa-onnx/csrc/speaker-embedding-extractor-nemo-model.h
+++ b/sherpa-onnx/csrc/speaker-embedding-extractor-nemo-model.h
@@ -6,11 +6,6 @@

 #include <memory>

-#if __ANDROID_API__ >= 9
-#include "android/asset_manager.h"
-#include "android/asset_manager_jni.h"
-#endif
-
 #include "onnxruntime_cxx_api.h"  // NOLINT
 #include "sherpa-onnx/csrc/speaker-embedding-extractor-nemo-model-meta-data.h"
 #include "sherpa-onnx/csrc/speaker-embedding-extractor.h"
@@ -22,10 +17,9 @@ class SpeakerEmbeddingExtractorNeMoModel {
  explicit SpeakerEmbeddingExtractorNeMoModel(
      const SpeakerEmbeddingExtractorConfig &config);

-#if __ANDROID_API__ >= 9
+  template <typename Manager>
  SpeakerEmbeddingExtractorNeMoModel(
-      AAssetManager *mgr, const SpeakerEmbeddingExtractorConfig &config);
-#endif
+      Manager *mgr, const SpeakerEmbeddingExtractorConfig &config);

  ~SpeakerEmbeddingExtractorNeMoModel();

--- a/sherpa-onnx/csrc/speaker-embedding-extractor.cc
+++ b/sherpa-onnx/csrc/speaker-embedding-extractor.cc
@@ -6,6 +6,15 @@

 #include <vector>

+#if __ANDROID_API__ >= 9
+#include "android/asset_manager.h"
+#include "android/asset_manager_jni.h"
+#endif
+
+#if __OHOS__
+#include "rawfile/raw_file_manager.h"
+#endif
+
 #include "sherpa-onnx/csrc/file-utils.h"
 #include "sherpa-onnx/csrc/macros.h"
 #include "sherpa-onnx/csrc/speaker-embedding-extractor-impl.h"
@@ -55,11 +64,10 @@ SpeakerEmbeddingExtractor::SpeakerEmbeddingExtractor(
    const SpeakerEmbeddingExtractorConfig &config)
    : impl_(SpeakerEmbeddingExtractorImpl::Create(config)) {}

-#if __ANDROID_API__ >= 9
+template <typename Manager>
 SpeakerEmbeddingExtractor::SpeakerEmbeddingExtractor(
-    AAssetManager *mgr, const SpeakerEmbeddingExtractorConfig &config)
+    Manager *mgr, const SpeakerEmbeddingExtractorConfig &config)
    : impl_(SpeakerEmbeddingExtractorImpl::Create(mgr, config)) {}
-#endif

 SpeakerEmbeddingExtractor::~SpeakerEmbeddingExtractor() = default;

@@ -77,4 +85,14 @@ std::vector<float> SpeakerEmbeddingExtractor::Compute(OnlineStream *s) const {
  return impl_->Compute(s);
 }

+#if __ANDROID_API__ >= 9
+template SpeakerEmbeddingExtractor::SpeakerEmbeddingExtractor(
+    AAssetManager *mgr, const SpeakerEmbeddingExtractorConfig &config);
+#endif
+
+#if __OHOS__
+template SpeakerEmbeddingExtractor::SpeakerEmbeddingExtractor(
+    NativeResourceManager *mgr, const SpeakerEmbeddingExtractorConfig &config);
+#endif
+
 }  // namespace sherpa_onnx
--- a/sherpa-onnx/csrc/speaker-embedding-extractor.h
+++ b/sherpa-onnx/csrc/speaker-embedding-extractor.h
@@ -9,11 +9,6 @@
 #include <string>
 #include <vector>

-#if __ANDROID_API__ >= 9
-#include "android/asset_manager.h"
-#include "android/asset_manager_jni.h"
-#endif
-
 #include "sherpa-onnx/csrc/online-stream.h"
 #include "sherpa-onnx/csrc/parse-options.h"

@@ -45,10 +40,9 @@ class SpeakerEmbeddingExtractor {
  explicit SpeakerEmbeddingExtractor(
      const SpeakerEmbeddingExtractorConfig &config);

-#if __ANDROID_API__ >= 9
-  SpeakerEmbeddingExtractor(AAssetManager *mgr,
+  template <typename Manager>
+  SpeakerEmbeddingExtractor(Manager *mgr,
                            const SpeakerEmbeddingExtractorConfig &config);
-#endif

  ~SpeakerEmbeddingExtractor();