quality_check.h

/* Copyright 2023 ID R&D Inc. All Rights Reserved. */
 
#pragma once
 
#include <voicesdk/core/config.h>
 
#include <iostream>
#include <vector>
#include <memory>
#include <unordered_map>
#include <string>
 
namespace voicesdk {
 
struct QualityCheckEngineResult {
 
    enum class QualityCheckShortDescription : uint8_t {
        kTooNoisy = 0,
 
        kTooSmallSpeechLength = 1,
 
        kOk = 2,
 
        kTooSmallSpeechRelativeLength = 3,
 
        kMultipleSpeakersDetected = 4
    };
 
    float snr_db;
 
    float speech_length_ms;
 
    float speech_relative_length;
 
    float multiple_speakers_detector_score;
 
    QualityCheckShortDescription quality_check_short_description;
 
    friend std::ostream& operator<<(std::ostream& os, const QualityCheckEngineResult& obj) {
        std::string short_description_str;
 
        switch (obj.quality_check_short_description) {
            case QualityCheckShortDescription::kTooNoisy:
                short_description_str = "kTooNoisy";
                break;
            case QualityCheckShortDescription::kTooSmallSpeechLength:
                short_description_str = "kTooSmallSpeechLength";
                break;
            case QualityCheckShortDescription::kOk:
                short_description_str = "kOk";
                break;
            case QualityCheckShortDescription::kTooSmallSpeechRelativeLength:
                short_description_str = "kTooSmallSpeechRelativeLength";
                break;
            case QualityCheckShortDescription::kMultipleSpeakersDetected:
                short_description_str = "kMultipleSpeakersDetected";
                break;
        }
 
        os << "QualityCheckEngineResult["
           << "snr_db: " << obj.snr_db << ", "
           << "speech_length_ms: " << obj.speech_length_ms << ", "
           << "speech_relative_length: " << obj.speech_relative_length << ", "
           << "multiple_speakers_detector_score: " << obj.multiple_speakers_detector_score << ", "
           << "quality_check_short_description: " << short_description_str
           << "]";
        return os;
    }
 
    bool operator==(const QualityCheckEngineResult& other) const {
        return quality_check_short_description == other.quality_check_short_description && snr_db == other.snr_db &&
               speech_length_ms == other.speech_length_ms && speech_relative_length == other.speech_relative_length &&
               multiple_speakers_detector_score == other.multiple_speakers_detector_score;
    }
};
 
struct QualityCheckMetricsThresholds {
    float minimum_snr_db = 0;
 
    float minimum_speech_length_ms = 0;
 
    float minimum_speech_relative_length = 0;
 
    float maximum_multiple_speakers_detector_score = 0;
};
 
enum class QualityCheckScenario: uint8_t {
    kVerifyTiEnrollment = 0,
 
    kVerifyTiVerification = 1,
 
    kVerifyTdEnrollment = 2,
 
    kVerifyTdVerification = 3,
 
    kLiveness = 4
};
 
class VOICE_SDK_API QualityCheckEngine {
public:
    using Ptr = std::shared_ptr<QualityCheckEngine>;
 
    static QualityCheckEngine::Ptr Create(const std::string& init_path);
 
    virtual ~QualityCheckEngine() = default;
 
    virtual QualityCheckMetricsThresholds GetRecommendedThresholds(QualityCheckScenario scenario) const = 0;
 
    virtual QualityCheckEngineResult CheckQuality(const std::string& audio_path,
                                                  const QualityCheckMetricsThresholds& thresholds) const = 0;
 
    virtual QualityCheckEngineResult CheckQuality(const float* float_samples, size_t num_samples, size_t sample_rate,
                                                  const QualityCheckMetricsThresholds& thresholds) const = 0;
 
    virtual QualityCheckEngineResult CheckQuality(const int16_t* pcm16_samples, size_t num_samples, size_t sample_rate,
                                                  const QualityCheckMetricsThresholds& threshold) const = 0;
 
    virtual QualityCheckEngineResult CheckQuality(const uint8_t* pcm16_bytes, size_t num_bytes, size_t sample_rate,
                                                  const QualityCheckMetricsThresholds& thresholds) const = 0;
};
 
}  // namespace voicesdk