Skip to content

feat: add audioOps mapper operators#461

Open
starlight6336 wants to merge 4 commits intoModelEngine-Group:mainfrom
starlight6336:main
Open

feat: add audioOps mapper operators#461
starlight6336 wants to merge 4 commits intoModelEngine-Group:mainfrom
starlight6336:main

Conversation

@starlight6336
Copy link
Copy Markdown

新增算子清单(15 个)
链路能力算子(pipeline steps)
audio_format_convert(AudioFormatConvert):常见音频格式互转 + 可选重采样/声道转换(输出到 export_path 并更新 filePath/fileType/fileName)
audio_anomaly_filter(AudioAnomalyFilter):异常语音检测与过滤(时长/静音比例),报告写入 ext_params.audio_quality
audio_gtcrn_denoise(AudioGtcrnDenoise):GTCRN ONNX 智能降噪(复用 audio_preprocessor 工具),输出到 export_path
audio_quantize_encode(AudioQuantizeEncode):重采样 + WAV PCM 8/16/24/32bit 量化编码(输出固定 .wav)
audio_fast_lang_id(AudioFastLangId):快速 LID(zh/en),结果写入 ext_params.audio_lid.lang
audio_asr_pipeline(AudioAsrPipeline):端到端流水线(normalization→可选降噪→可选异常过滤→LID→切分→ASR→合并),最终写入 sample["text"]
行业 DSP 算子(可叠加组合)
audio_dc_offset_removal(AudioDcOffsetRemoval):去直流分量(减均值)
audio_pre_emphasis(AudioPreEmphasis):预加重
audio_simple_agc(AudioSimpleAgc):分段 RMS 自动增益(AGC)
audio_noise_gate(AudioNoiseGate):噪声门
audio_soft_peak_limiter(AudioSoftPeakLimiter):软限幅(tanh 近似)
audio_trim_silence_edges(AudioTrimSilenceEdges):首尾静音裁剪(带 padding)
audio_rms_loudness_normalize(AudioRmsLoudnessNormalize):整段 RMS 归一 + 峰值顶限
audio_hum_notch(AudioHumNotch):50/60Hz 工频陷波(依赖 scipy.signal)
audio_telephony_bandpass(AudioTelephonyBandpass):电话带通 300–3400Hz(依赖 scipy.signal)
Implementation notes
已更新 DataMate/runtime/ops/mapper/init.py,新增上述算子包的 import,保证启动加载与 OPERATORS.register_module(...) 注册生效。

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant