bin/test_silero.cpp

#include <iostream>
#include <vector>
#include <cmath>
#include <iomanip>      // std::fixed, std::setprecision
#include <fstream>      // std::ifstream
#include <string>       // std::string

// 自定义头文件
#include "wav.h"         // 包含 wav::WavReader 定义
#include "time_stamp.h"          // 包含 timestamp_t 定义
#include "vad_iterator.h"       // 包含 VadIterator 类声明

int main(int argc, char* argv[]) {
    if (argc < 3) {
        std::cerr << "Usage: " << argv[0] << " <model_absolute_path>" << " <audio_list_absolute_path>" << std::endl;
        return 1;
    }

    // 获取命令行传入的模型路径和音频列表文件路径
    std::string model_path = argv[1];
    std::string audio_list_path = argv[2];

    // 打开 audio_list.txt 文件
    std::ifstream audio_list_file(audio_list_path);
    if (!audio_list_file.is_open()) {
        std::cerr << "Error: Unable to open audio list file: " << audio_list_path << std::endl;
        return 1;
    }

    // 初始化 VadIterator
    VadIterator vad(model_path);

    // 逐行读取音频文件路径并处理
    std::string wav_path;
    while (std::getline(audio_list_file, wav_path)) {
        if (wav_path.empty()) {
            continue; // 跳过空行
        }

        std::cout << wav_path << std::endl;

        try {
            // 读取 WAV 文件 (expects 16000 Hz, mono, PCM)
            wav::WavReader wav_reader(wav_path);
            int numSamples = wav_reader.num_samples();
            std::vector<float> input_wav(static_cast<size_t>(numSamples));
            for (size_t i = 0; i < static_cast<size_t>(numSamples); i++) {
                input_wav[i] = static_cast<float>(*(wav_reader.data() + i));
            }

            // 处理音频
            vad.process(input_wav);
            // 获取语音时间戳 (以样本为单位)
            std::vector<timestamp_t> stamps = vad.get_speech_timestamps();

            // 将时间戳转换为秒并输出
            const float sample_rate_float = 16000.0f;
            for (size_t i = 0; i < stamps.size(); i++) {
                float start_sec = std::rint((stamps[i].start / sample_rate_float) * 10.0f) / 10.0f;
                float end_sec = std::rint((stamps[i].end / sample_rate_float) * 10.0f) / 10.0f;
                // std::cout << "Speech detected from "
                //           << std::fixed << std::setprecision(1) << start_sec
                //           << " s to "
                //           << std::fixed << std::setprecision(1) << end_sec
                //           << " s"
                //           << " [ " << stamps[i].start << " " << stamps[i].end << " ]"
                //           << std::endl;
                std::cout << stamps[i].start << ", " << stamps[i].end << std::endl;
            }

            // 重置内部状态
            vad.reset();
        } catch (const std::exception& e) {
            std::cerr << "Error processing file " << wav_path << ": " << e.what() << std::endl;
        }
    }

    audio_list_file.close();
    return 0;
}