菜单
Use when performing FFT, STFT, Welch PSD estimation, 1/N octave band analysis, coherence, cross-spectral density, or transfer function analysis with wandas.
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
基于 SOC 职业分类
| name | wandas-spectral-analysis |
| description | Use when performing FFT, STFT, Welch PSD estimation, 1/N octave band analysis, coherence, cross-spectral density, or transfer function analysis with wandas. |
ChannelFrame を周波数域フレームに変換する。各メソッドは新しいフレームを返す。
np.fft.fft() 等の直接呼び出し禁止。必ず wandas メソッドを使う。.plot() / .describe() を使う。plt.plot は禁止。| メソッド | シグネチャ | 返り値 |
|---|---|---|
.fft | (n_fft: int = None, window: str = "hann") | SpectralFrame |
.welch | (n_fft: int = 2048, hop_length: int = None, win_length: int = None, window: str = "hann", average: str = "mean") | SpectralFrame |
.stft | (n_fft: int = 2048, hop_length: int = None, win_length: int = None, window: str = "hann") | SpectrogramFrame |
.noct_spectrum | (fmin: float = 25, fmax: float = 20000, n: int = 3, G: int = 10, fr: int = 1000) | NOctFrame |
.coherence | (n_fft: int = 2048, hop_length: int = None, win_length: int = None, window: str = "hann", detrend: str = "constant") | SpectralFrame |
.csd | (n_fft: int = 2048, hop_length: int = None, win_length: int = None, window: str = "hann", detrend: str = "constant", scaling: str = "spectrum", average: str = "mean") | SpectralFrame |
.transfer_function | (n_fft: int = 2048, hop_length: int = None, win_length: int = None, window: str = "hann", detrend: str = "constant", scaling: str = "spectrum", average: str = "mean") | SpectralFrame |
デフォルト値: hop_length = n_fft // 4、win_length = n_fft（None 時）
| メソッド | 入力フレーム | 返り値 |
|---|---|---|
.ifft() | SpectralFrame | ChannelFrame |
.istft() | SpectrogramFrame | ChannelFrame |
| プロパティ | 説明 |
|---|---|
.freqs | 周波数軸（Hz）の ndarray |
.magnitude | 振幅スペクトル |
.phase | 位相スペクトル |
.power | パワースペクトル |
.dB | dB 換算値 |
.dBA | A 重み付き dB |
| プロパティ | 説明 |
|---|---|
.freqs | 周波数軸（Hz） |
.times | 時間軸（秒） |
.n_frames | 時間フレーム数 |
.n_freq_bins | 周波数ビン数 |
.magnitude | 振幅スペクトログラム |
.dB | dB スペクトログラム |
.dBA | A 重み付き dB スペクトログラム |
import wandas as wd
signal = wd.read_wav("audio.wav")
spectrum = signal.fft()
spectrum.plot(title="Frequency Spectrum")
print(f"卓越周波数: {spectrum.freqs[spectrum.data.argmax()]:.1f} Hz")
psd = signal.welch(n_fft=2048, hop_length=512)
psd.plot(title="Power Spectral Density")
import wandas as wd
sensor = wd.read_csv("sensor.csv", time_column="Time")
spectrogram = (sensor
.high_pass_filter(cutoff=50)
.stft(n_fft=2048, hop_length=512))
spectrogram.plot(cmap="inferno", title="Anomaly Detection")
noct = signal.noct_spectrum(fmin=25, fmax=8000, n=3)
noct.plot(title="1/3 Octave Band Spectrum")
import wandas as wd
ref = wd.read_wav("ref.wav")
meas = wd.read_wav("meas.wav")
combined = ref.add_channel(meas)
coh = combined.coherence(n_fft=2048)
coh.plot(title="Coherence")
combined = input_sig.add_channel(output_sig)
tf = combined.transfer_function(n_fft=2048)
tf.plot_matrix() # チャンネル間マトリクス表示
| 間違い | 正解 |
|---|---|
SpectralFrame.fft() を呼ぶ | すでに周波数域。ifft() で時間域に戻す |
welch() が SpectrogramFrame と思う | welch() は SpectralFrame（時間平均済み）を返す |
.coherence() を単chフレームで呼ぶ | コヒーレンスは 2ch 以上必要 |
noct_spectrum の fmax がナイキスト以下と思う | fmax デフォルトは 20000（sr 非依存） |
welch の average を忘れる | average="mean"（デフォルト）または "median" |
examples/workflows.md — スペクトル解析のシナリオ別コード例references/spectral_api.md — 全変換メソッドの詳細 APIUse when auditing screenshots or live UI for cognitive load, usability friction, and heuristic violations.
Use when reproducing or preventing runtime-only Webview UI bugs with real Chromium checks, Playwright smoke specs, static HTML-pattern linting, or dogfooding the UI smoke layers against a suspected regression.
Use when analyzing audio or vibration signals end-to-end, generating Jupyter Notebook analysis reports, comparing multiple measurement conditions, evaluating noise or sound quality, detecting anomalies in sensor data, or performing adaptive signal investigation driven by hypothesis and findings with wandas.
Use when starting with wandas, loading audio or sensor data from WAV or CSV files, creating signals from NumPy arrays, understanding ChannelFrame and other frame types, inspecting signal metadata, or setting up physical units (Pa, m/s²) for dB calculations.
Use when applying filters (lowpass, highpass, bandpass, A-weighting), normalizing signals, resampling, trimming, adding fades, computing RMS trends, calculating sound level (dB, A-weighting), or computing psychoacoustic metrics (loudness, roughness, sharpness) with wandas.
Use when plotting waveforms, frequency spectra, spectrograms, octave band charts, roughness heatmaps, overlaying multiple signals on the same axes, or configuring describe() with frequency range and colormap settings with wandas.