Help

Workspace

Analysis sidebar: Time: Env, Time: F0, Freq: Spec, Freq: Ridge and Disp: Show mini-tabs provide one editable location for analysis controls. The sidebar may be collapsed; on narrow screens it behaves as a drawer. The fixed footer supplies contextual zoom navigation. The strip below the displays is read-only and links back to the related mini-tab.

Analysis workflow

Mode?: After WAV files are loaded, choose Parameter Setup or Exploratory Flex. Comparison Batch becomes available after Parameter Setup.

Parameter Setup: Opens the first file as an orientation preview and lets you adjust one shared configuration. Select Comparison Batch from the same dropdown to start the closed batch.

Comparison Batch: Shows the complete F0 profile summary for confirmation, then processes every file automatically with frozen settings, exports the completed batch and opens SCATTER. ABORT discards all interim batch results.

Exploratory Flex: Opens the first file using the defaults and preserves free per-file analysis and zooming without SCATTER.

Toolbar

Choose WAV files: Select WAV files. All decodable WAV files are accepted; filenames are used only for metadata. Categories are inferred from filenames of the form <categoryToken>_<anything>.wav or <categoryToken>-<anything>.wav and displayed as compact 1- or 2-character codes.

Balanced sampling: Choose one or more categories in the top bar, enter the requested number per category and a reproducible seed, then select SAMPLE. Sampling is without replacement. If any selected category contains fewer files than requested, every selected category is reduced to the size of the smallest selected category. RESAMPLE increments the seed; ALL FILES restores the complete loaded source database.

◀ PREV: Return to the previous file.

NEXT ▶: Process/display the next file.

File chip: Shows the current file number, e.g. 3/10.

EXP: Export ZIP.

📊 SCATTER: Scatter plots.

k-means popup: Opens an independent three-column teaching workspace with settings, a square plot and scrollable diagnostics. k-means is unsupervised: known categories do not determine the clusters. ProsCanvas reports silhouette and variation measures, contingency counts with row and column percentages, adjusted Rand correspondence, case-level assignments and explicit limitations.

Square ML plots: Linear SVM and k-means popup charts use square plotting areas so horizontal and vertical visual distances are not distorted by the panel aspect ratio.

Full-width SCATTER workspace: SCATTER replaces the ordinary sidebar/display workspace below the top command row, so the plots and Results use the complete available width. Closing SCATTER restores the previous sidebar tab, scroll position, zoom histories and information strip.

Sidebar zoom footer: Back, Forward, Zoom to selection, Full range and Clear operate on the active time or spectrum domain. Time, ENV-spectrum and F0-spectrum histories remain independent; Full range is navigable and does not erase history.

↺ RESET: Factory reset.

Disp: Show tab: Upper- and lower-display visibility controls are located in the left sidebar and repaint only; they do not alter numerical analysis or SCATTER operations.

F0 display

F0: Switches the time canvas from the standard waveform + envelope display to a waveform hint + F0 display. In F0 mode the button is labelled ENV and switches back to the standard display.

The F0 procedure uses the F0 Explorer v1.1.6-compatible engine: DC-centred and peak-normalised input, anti-aliased analysis-rate conversion (8 kHz factory setting), a Butterworth high-pass/low-pass cascade, Tukey-windowed FFT-YIN, Central 75% automatic range estimation, explicit voicing and silence decisions, bounded silence-aware bridging, and gap-aware smoothing. Originally voiced F0 is drawn solid blue; accepted short bridges are lighter blue and dotted; longer or silence-associated gaps remain missing. The compact F0 controls expose band-pass limits and smoothing window. Advanced F0… opens the visually secondary Dev Mode with the complete calculation settings, four independent persistent Male, Female, Both and Custom settings domains, domain-specific Factory restoration, production-profile promotion, and JSON profile import/export. Advanced edits remain in the selected domain and invalidate dependent results.

Frequency domain

Spec:gram / Spectrum, Window and Hops/win: The conspicuous action button switches between the static spectrum and an ENV or F0 time-frequency spectrogram; its text names the view that will open. Spectrogram frequency uses the active spectrum range; spectrogram time follows the upper-canvas time range. Window choices are 0.5 s, 1 s, and then 2–20 s in 1 s steps; the default remains 4 s. Hop duration is window duration divided by hops per window. Colours use YlOrRd from pale yellow (low magnitude) to dark red (high magnitude).

TRACKS: Builds a ridge-evidence image from the ENV or F0 spectrogram, detects elongated object bodies, extracts centre lines constrained to those bodies, overlays up to three principal objects, and opens aligned frequency and associated global-magnitude plots. Magnitude prominence may cross without changing object identity.

ENV and F0 spectra support raw, MOVWIN-smoothed, and kernel-smoothed peak-evidence views. The v8 gap-aware downstream analysis preserves retained F0 gaps: contiguous finite runs are transformed separately and combined in the power domain with valid-duration weighting. F0 spectrogram windows, RFA candidates and objects are calculated only where finite-run coverage is sufficient; unsupported windows remain explicitly unavailable rather than appearing as zero energy. Fuzzy shaded bands indicate approximate RFA reference zones around syllable-pair, syllable-rate, and fast-syllable edge regions.

The default RFA range is 1–6 Hz with conservative DC removal at 0.1 Hz. Shaded bands are RFA reference zones, not measured evidence. The peak-zone curve is kernel-smoothed evidence from spectral peaks. Candidate rhythm-formant zones combine peak clustering, prior overlap, and temporal plausibility; they are candidates only, not final rhythm identifications.

The LF spectrum shows the raw low-frequency envelope-spectrum magnitudes. The RFA-smoothed LF spectrum suppresses narrow bin-to-bin fluctuations and highlights broader rhythm-formant regions. Candidate RFA peaks and zones should be interpreted with this smoothing relationship in mind: the smoothed curve clarifies the LF spectrum but does not constitute a second independent measurement.

PWR: When enabled for the active mode, FFT magnitudes are squared before normalisation, MOVWIN smoothing, peak detection, kernel evidence, and RFA candidate processing. This is a squared-magnitude display, not a calibrated power spectral density. ENV and F0 retain separate PWR settings.

MOVWIN: MA uses an equal-weight moving average, GAU uses Gaussian weights, HAN uses a Hann taper, and TUK uses a Tukey taper with selectable α; α = 0 is rectangular. These methods smooth neighbouring frequency bins only; the envelope low-pass cutoff is a separate time-domain smoothing control.

SPEC / LPF / PEAK: SPEC shows or hides the raw green spectrum, LPF shows or hides the blue low-pass-smoothed spectrum, and PEAK shows or hides spectral peak markers and peak information. These display controls are independent.

MOVWIN bins: selects the moving-window length in spectrum bins: 3, 5, 7, 9, 11, 13, or 15. If the spectrum of the selected envelope has fewer than twice the selected MOVWIN bin length, MOVWIN is not applied and a warning is shown.

Event Rhythm measures

IPU: interpausal unit (traditionally “run”). Initial and final silence are excluded; qualifying internal silence regions are pauses.

Event Rhythm vector representations: SCATTER provides proportional peak-time, maximum-normalised peak-magnitude, raw GLOB inter-peak-interval and mean-normalised GLOB inter-peak-interval trajectories. They use the canonical full-analysis ENV peaks. Raw intervals preserve tempo and pattern; mean-normalised intervals remove mean tempo. These sequences underlie GLOB timing analysis but are not themselves nPVI values.

IPUs with fewer than three peaks do not contribute zero or NaN to InPVI; they are excluded before averaging. Undefined displayed values are shown as —.

Scatter machine learning

SCATTER command rail: Use the compact Regression and Analysis selectors for the active display. SETTINGS opens a contextual drawer containing only the controls relevant to the selected method. DONE, the drawer close button, Escape, or the shaded backdrop closes the drawer without closing SCATTER.

Chart font size: The 8–24 px control at the bottom of the SCATTER panel changes the shared typography of every SCATTER chart. To compensate provisionally for popup-canvas display scaling, the Linear SVM and k-means popup charts render at the selected base size plus 2 pt (2.667 px). It redraws presentation only; analytical results are not recalculated. PNG output uses the displayed main-chart size.

Linear SVM popup: OPEN LINEAR SVM opens a large independent working area for pair choice, category or k-means target, equal-case or equal-category weighting, fixed or Auto C, canonical or balanced-threshold boundary, validation controls, enlarged plot, diagnostics and CSV exports. Popup edits remain unapplied until APPLY TO SCATTER; Cancel or close discards them. Auto C uses the declared deterministic grid and balanced-accuracy tie rules. Balanced-threshold mode shifts the fitted direction parallel and is explicitly distinct from the canonical maximum-margin boundary. Held-out evaluation fits scaling, C selection and threshold selection within training data only.

Vector eligibility guidance: If fewer than two valid vectors remain, ProsCanvas reports the selected representation and distance, eligible and excluded counts, and directs the learner to Excluded recordings. For ENV/F0 spectral vectors, Exploratory Flex may be used to inspect the relevant spectrogram tracks. For Event Rhythm vectors, inspect the ENV waveform, envelope and detected peak markers for too few or unsuitable peaks and intervals. This is an eligibility message, not a judgement about the scientific adequacy of the full dataset.

Variable picker: Select either exactly three scalar variables or one vector representation from the eight colour-coded empirical-source groups. Each row shows the variable name, empirical source and unit; search filters names, sources, definitions and units. DESELECT ALL clears the complete scalar draft selection, including variables hidden in collapsed groups or by search; select exactly three before applying. Scalar selections are A, B and C, producing A × B, A × C and B × C. Vector variables are selected one at a time from separate ENV, F0 and Event Rhythm groups and produce an independent hierarchical dendrogram with the selected distance and linkage settings. F0 vectors use the smoothed contour with internal gaps linearly interpolated before FFT analysis; leading and trailing gaps are not extrapolated. Unsupported values remain unavailable and are excluded rather than converted to zero.

Plain: category-coloured scatter with file-number point labels.

k-means: unsupervised clustering in the selected A, B and C variables after internal z-standardisation. Colours show K-clusters; labels show category plus file number.

Linear SVM: binary supervised boundary on the displayed two-feature scatter axes. The main plot shows the explicitly applied full-data model. Canonical mode uses decision score zero; balanced-threshold mode retains the fitted direction and moves only the parallel offset. The popup reports balanced and overall accuracy, category recalls, separate fitted-sample and held-out confusion matrices, explicit TP/TN/FP/FN orientation, support vectors, margin violations and category score dispersion. A teaching interpretation explains what can be inferred positively, what the analysis does not establish, and ranks the three variable pairs by held-out balanced accuracy where available. Training fit is descriptive; leave-one-out and repeated stratified cross-validation estimate held-out performance. Save CSV and Export preserve configuration, coefficients, thresholds, diagnostics, matrices, interpretation and case predictions.

Kernel SVM: binary RBF-kernel SVM on the displayed two-feature scatter axes. The solid contour is the non-linear decision boundary; dotted contours are approximate margin contours. C and gamma are compact controls. Fit agreement is descriptive and is computed on displayed training data.

DENDRO: three one-variable horizontal dendrograms for the selected A, B and C variables. DENDRO mode is exploratory teaching/study output and does not validate category labels or k-means clusters independently. For one-variable dendrograms, Manhattan and Euclidean distances both reduce to absolute difference; linkage choice usually has the larger visible effect. Ward linkage is provided only with Euclidean distance.

For SVM target K-means: The SVM - k-means comparison is exploratory only and does not imply validation.