OmniX-Space/xiaozhi-esp32
1
0
Fork 0
Code Issues Pull Requests Actions Releases Activity

modify AddAudioData

Browse Source
This commit is contained in:
苏雨洛
2025-09-10 17:29:05 +08:00
parent ed9fb33f80
commit 50771c571d
4 changed files with 114 additions and 128 deletions
Download Patch File Download Diff File Expand all files Collapse all files

91
main/application.cc
View File

@@ -794,9 +794,6 @@ void Application::AddAudioData(AudioStreamPacket&& packet) {
// 检查采样率是否匹配，如果不匹配则进行简单重采样
if (packet.sample_rate != codec->output_sample_rate()) {
// ESP_LOGI(TAG, "Resampling music audio from %d to %d Hz",
// packet.sample_rate, codec->output_sample_rate());
// 验证采样率参数
if (packet.sample_rate <= 0 || codec->output_sample_rate() <= 0) {
ESP_LOGE(TAG, "Invalid sample rates: %d -> %d",
@@ -804,51 +801,61 @@ void Application::AddAudioData(AudioStreamPacket&& packet) {
return;
}
std::vector<int16_t> resampled;
if (packet.sample_rate > codec->output_sample_rate()) {
ESP_LOGI(TAG, "Music playback: Adjust the sampling rate from %d Hz switch to %d Hz",
codec->output_sample_rate(), packet.sample_rate);
// 尝试动态切换采样率
if (codec->SetOutputSampleRate(packet.sample_rate)) {
ESP_LOGI(TAG, "Successfully switched to music playback sampling rate: %d Hz", packet.sample_rate);
} else {
ESP_LOGW(TAG, "Unable to switch sampling rate, continue using current sampling rate: %d Hz", codec->output_sample_rate());
}
// 尝试动态切换采样率
if (codec->SetOutputSampleRate(packet.sample_rate)) {
ESP_LOGI(TAG, "Successfully switched to music playback sampling rate: %d Hz", packet.sample_rate);
} else {
// 上采样：线性插值
float upsample_ratio = codec->output_sample_rate() / static_cast<float>(packet.sample_rate);
size_t expected_size = static_cast<size_t>(pcm_data.size() * upsample_ratio + 0.5f);
resampled.reserve(expected_size);
for (size_t i = 0; i < pcm_data.size(); ++i) {
// 添加原始样本
resampled.push_back(pcm_data[i]);
ESP_LOGW(TAG, "Unable to switch sampling rate, continue using current sampling rate: %d Hz", codec->output_sample_rate());
// 如果无法切换采样率，继续使用当前的采样率进行处理
if (packet.sample_rate > codec->output_sample_rate()) {
// 下采样：简单丢弃部分样本
float downsample_ratio = static_cast<float>(packet.sample_rate) / codec->output_sample_rate();
size_t expected_size = static_cast<size_t>(pcm_data.size() / downsample_ratio + 0.5f);
std::vector<int16_t> resampled(expected_size);
size_t resampled_index = 0;
// 计算需要插值的样本数
int interpolation_count = static_cast<int>(upsample_ratio) - 1;
if (interpolation_count > 0 && i + 1 < pcm_data.size()) {
int16_t current = pcm_data[i];
int16_t next = pcm_data[i + 1];
for (int j = 1; j <= interpolation_count; ++j) {
float t = static_cast<float>(j) / (interpolation_count + 1);
int16_t interpolated = static_cast<int16_t>(current + (next - current) * t);
resampled.push_back(interpolated);
}
} else if (interpolation_count > 0) {
// 最后一个样本，直接重复
for (int j = 1; j <= interpolation_count; ++j) {
resampled.push_back(pcm_data[i]);
for (size_t i = 0; i < pcm_data.size(); ++i) {
if (i % static_cast<size_t>(downsample_ratio) == 0) {
resampled[resampled_index++] = pcm_data[i];
}
}
pcm_data = std::move(resampled);
ESP_LOGI(TAG, "Downsampled %d -> %d samples (ratio: %.2f)",
pcm_data.size(), resampled.size(), downsample_ratio);
} else if (packet.sample_rate < codec->output_sample_rate()) {
// 上采样：线性插值
float upsample_ratio = codec->output_sample_rate() / static_cast<float>(packet.sample_rate);
size_t expected_size = static_cast<size_t>(pcm_data.size() * upsample_ratio + 0.5f);
std::vector<int16_t> resampled(expected_size);
for (size_t i = 0; i < pcm_data.size(); ++i) {
// 添加原始样本
resampled[i * static_cast<size_t>(upsample_ratio)] = pcm_data[i];
// 计算需要插值的样本数
int interpolation_count = static_cast<int>(upsample_ratio) - 1;
if (interpolation_count > 0 && i + 1 < pcm_data.size()) {
int16_t current = pcm_data[i];
int16_t next = pcm_data[i + 1];
for (int j = 1; j <= interpolation_count; ++j) {
float t = static_cast<float>(j) / (interpolation_count + 1);
int16_t interpolated = static_cast<int16_t>(current + (next - current) * t);
resampled[i * static_cast<size_t>(upsample_ratio) + j] = interpolated;
}
} else if (interpolation_count > 0) {
// 最后一个样本，直接重复
for (int j = 1; j <= interpolation_count; ++j) {
resampled[i * static_cast<size_t>(upsample_ratio) + j] = pcm_data[i];
}
}
}
pcm_data = std::move(resampled);
ESP_LOGI(TAG, "Upsampled %d -> %d samples (ratio: %.2f)",
pcm_data.size() / static_cast<size_t>(upsample_ratio), pcm_data.size(), upsample_ratio);
}
ESP_LOGI(TAG, "Upsampled %d -> %d samples (ratio: %.2f)",
pcm_data.size(), resampled.size(), upsample_ratio);
}
pcm_data = std::move(resampled);
}
// 确保音频输出已启用