Technical Operations Manual
This manual is the primary field reference for the SPECTER Paranormal Imaging System. It covers all operational procedures, interface elements, detection methodologies, and field protocols required for professional paranormal investigation. Operators are expected to read this document in full prior to first deployment. Familiarity with the system's detection architecture is essential for accurate interpretation of evidence.
System Overview
SPECTER is a multi-sensor spatial imaging platform purpose-built for paranormal field investigation. It fuses data from three simultaneous sensor channels, wide-spectrum RGB imaging, structured-light depth arrays, and inertial measurement unit (IMU) telemetry, against a continuously recalculated Environmental Anomaly Index. When that index drifts past a calibrated threshold, the system fires a DETECTION EVENT and autonomously archives a timestamped evidence package.
Unlike legacy SLS systems which passively display a skeleton overlay on a single depth feed, SPECTER is an active detection architecture. It does not require the operator to visually monitor the screen at all times. The system is watching, logging, comparing, and scoring the environment continuously, whether or not an investigator is looking at the display.
SPECTER operates in two primary modes:
All six environmental metrics active. The system performs continuous frame-by-frame analysis against a dynamic baseline, monitoring for luminance shifts, spectral changes, flow disruption, field coherence collapse, and baseline deviation. Primary mode for smoke-machine investigations, general sweeps, and multi-metric environmental monitoring.
Neural network entity tracking layer activated. Structured-light depth data feeds a keypoint detection model trained to identify humanoid forms. Three additional right-rail metrics become active: Subjects Tracked, Average Confidence, and Kinematic Coherence. Use when investigating locations with a history of apparition sightings or intelligent entity activity.
Primary Interface Anatomy
The SPECTER interface is partitioned into functional operational regions. Every element visible on screen carries live data, there are no decorative indicators. Understanding the layout before deployment is critical to real-time field interpretation.
2.1 Top Bar & Detector Control
The top bar provides persistent system status across all modes and views. From left to right:
| ELEMENT | DESCRIPTION |
|---|---|
| System Glyph | SPECTER identifier mark. Pulses amber during active detection events. |
| Active Source Label | Identifies the connected sensor: REALSENSE D435i, KINECT V2, or HDMI CAPTURE. Flashes red if sensor connection is interrupted. |
| MODE | Current operating mode: ANOMALY or SLS. Tap to toggle. |
| VIEW | Current visual overlay: RGB (full color) or DEPTH (false-color depth map). Depth view renders entity outlines more clearly in low-light environments. |
| ENV PRESET | Active environmental preset: NEUTRAL, FOG, LOW LIGHT, or OPEN AIR. Adjusts baseline weighting for the environmental medium in use. |
| CLASS | Detection sensitivity class: CLASS-I (40%), CLASS-II (60%), CLASS-III (80%). Sets the Anomaly Index threshold that triggers a logged detection event. |
| Frame Counter | Live frame count for the current session. Used to cross-reference detection events with external recording timestamps. |
2.2 Detector Array Chips
Four status chips sit in the detector bar below the top bar. Each chip corresponds to a distinct detection subsystem. A chip glows cyan when that detector array is armed and feeding data to the Anomaly Index. Tap any chip to toggle that channel on or off without interrupting the session.
| CHIP | SUBSYSTEM | WHAT IT MONITORS |
|---|---|---|
| FACE | Facial Landmark Detection | Neural network scans the RGB frame for facial geometry, both visible and partial occlusions. Triggers on confident detections above confidence threshold. |
| BODY | Full-Body Pose Estimation | Full-frame pose estimation model. Distinct from SLS Mode, operates on the RGB channel and does not require depth data. Lower false-positive rate in well-lit environments. |
| SPECTRAL | Spectral Density Monitor | Shannon entropy analysis of the RGB histogram. Flags sudden collapse or expansion of spectral distribution, indicating something with a distinct chromatic signature entered the frame. |
| SLS | Structured Light Sensor | Arms the full SLS neural tracking pipeline. Requires depth sensor hardware. Activates the right-rail trio of SLS-specific metrics. |
System Meters, Left Rail
The left rail displays six live instrument readouts with continuous environmental updates at the camera's native frame rate. Each meter feeds directly into the Anomaly Index composite score. All metrics are derived from real sensor physics, these are not arbitrary indicators. Understanding the technical basis of each meter is essential for accurate evidence interpretation.
Technical basis: Mean brightness of the current frame expressed as a value from 0 to 255, plus the delta from a continuously updated exponentially-weighted moving average. "cd/m²" (candela per square metre) is a standard photometric unit measuring luminous intensity per unit area, literally the number of photons striking the camera sensor per frame interval.
What the system is measuring: Any change in the ambient light energy reaching the sensor. This includes light source changes, physical occlusion of the lens, and spectral shifts invisible to the human eye in near-infrared bands.
- Sudden drop of 5+ below baseline, something occluded the lens or passed between the sensor and a light source. Investigate immediately.
- Sudden spike, a brief high-intensity light source appeared in frame. May correlate with reported flash phenomena.
- Sustained upward drift, room ambient lighting is genuinely changing, not anomalous. Cross-reference with ENV preset.
- Rapid oscillation around baseline, motion in the environment. Normal unless paired with elevated Anomaly Index.
Technical basis: Shannon entropy of the RGB histogram, averaged across red, green, and blue channels. Entropy (symbol: H) is the information-theoretic measure of disorder, a high H value means colors are widely distributed across the spectrum, a low H value means the frame is dominated by a narrow color range.
What the system is measuring: The chromatic complexity of the scene. A smoke-filled room has moderate entropy. A blank white wall has near-zero entropy. An entity with a distinct spectral signature, whether visible or in near-infrared, will shift entropy measurably when it enters the frame.
- Sudden entropy collapse, a large, spectrally uniform object entered the frame. In a clear room, this is significant. In a smoke investigation, this may indicate smoke thickening.
- Sudden entropy expansion, something with a different color palette appeared. A face. A translucent form with a shifted chromatic profile.
- Steady-state oscillation, smoke drift. Monitor but do not flag without corroboration from other meters.
Technical basis: Dense optical flow computed across every pixel block in the frame. For each block, the system calculates displacement between the current frame and the previous frame, producing a vector field of motion across the entire image. The dominant direction (0° = right, clockwise) and magnitude (pixels per frame) are reported in real time.
What the system is measuring: The direction and speed of movement in the environmental medium, typically atmospheric particulate in a fog investigation. In a smoke-free room, this measures any movement of objects within the frame. This metric is the system's primary detector of movement that bypasses the visible spectrum, because it measures displacement in the image plane, not light emission.
- Direction stable for 5+ seconds then rotates 90° or more, something disturbed the directional flow. High significance.
- Magnitude doubles or more without corresponding visual event, something moved through the medium without being visually apparent.
- Gradual magnitude increase, HVAC or natural air current strengthening. Check environmental controls.
Technical basis: The variance of the optical flow vector field, how aligned all local motion vectors are across the frame. A value of 1.0 (σ = 1.0) means every pixel block is moving in perfect lockstep. A value of 0.0 means pure directional chaos. Sigma (σ) denotes standard deviation, used here as the inverse-variance coherence score.
What the system is measuring: The structural integrity of the environmental medium's movement. A clean, undisturbed smoke cloud drifts as a single coherent body, high coherence. When a physical or non-physical object moves through it, the smoke separates around that object, creating discontinuous motion vectors, coherence collapses.
This is the system's most sensitive detector of entity presence in a fog-medium investigation. An entity that displaces atmospheric particulate without being visually registered will cause a measurable coherence drop even if it triggers no other metric.
- Drop below 0.5σ sustained for 0.5 seconds or more, primary disturbance threshold. Something is genuinely disrupting the flow field. This is the highest-weight trigger for the Anomaly Index.
- Brief dip and recovery, investigator movement, breath, or equipment vibration. Cross-reference with session log timestamps.
- Sustained collapse toward 0.0, ongoing disturbance. Something is actively present in the medium. Escalate to full evidence capture protocol.
Technical basis: L2 distance (Euclidean pixel-wise difference) between the current frame and the exponentially-weighted moving average of recent frames. Delta (Δ) denotes change from baseline. The baseline is a live, continuously updated representation of what "normal" looks like in the current environment, not a static reference frame captured at session start.
What the system is measuring: How different right now is from the established recent norm. This is the system's persistence detector, it distinguishes between transient events (a person walking past) and sustained anomalies (something that entered the environment and stayed).
- Spike followed by rapid return to baseline, transient event. Something flashed through or briefly appeared.
- Spike followed by sustained elevation, something new entered the scene and is persisting. High significance. Log all other metrics at this timestamp.
- Gradual upward drift, environmental change over time (temperature shift, lighting change). Not anomalous in isolation.
- LOG DISPLAYS "BASELINE NOMINAL", deviation within expected variance. Environment is stable and calibrated.
Technical basis: Composite confidence score. Weighted blend of three primary contributors:
The candidate region score is the highest-scoring sub-region within the frame, the system identifies candidate areas and scores each one, taking the maximum. Inverse field coherence converts the coherence drop into a positive contribution score. Normalized baseline deviation scales the raw L2 distance into a 0–1 range for weighting.
What the system is measuring: A unified, single-number answer to the question: "Is something happening right now?" This is the primary operational readout. All other meters provide diagnostic context, the Anomaly Index is the field decision signal.
- CLASS-I 40%, Detection event fired. Low-confidence anomaly. Archive entry created. Continue monitoring.
- CLASS-II 60%, Strong anomaly. Archive entry created with depth map badge if depth sensor active. Recommend full investigator attention.
- CLASS-III 80%, Exceptional anomaly. Full evidence package archived. All active metrics snapshotted. Recommend immediate investigation of flagged region.
The threshold for detection event firing can be adjusted in CONFIG. Default is CLASS-I (40%). Raising the threshold in high-activity environments prevents archive saturation. Lowering it in quiet environments improves detection sensitivity for subtle manifestations.
SLS Mode, Neural Entity Tracking
SLS Mode activates SPECTER's structured-light entity detection pipeline. In this mode, depth data from the active sensor feeds a neural network keypoint detection model, which maps humanoid skeletal geometry in real time. Three dedicated right-rail metrics become active, providing granular intelligence on detected figures beyond a simple "body present / not present" read.
SLS Mode is designed to do what the original Kinect SLS camera could not: distinguish genuine anomalies from false detections. The Kinematic Coherence scoring system was built specifically to address the community's most persistent frustration with legacy SLS systems, chairs, curtains, coat racks, and investigators' own limbs generating false stick figures. In SPECTER, a detected figure that doesn't exhibit movement patterns consistent with a real entity is scored accordingly and will not escalate to a detection event.
Technical basis: Number of distinct humanoid figures currently detected by the neural network with at least one valid keypoint above the minimum confidence threshold. Each figure is assigned a persistent tracking ID for the duration of the session, allowing the system to distinguish between the same entity reappearing and multiple simultaneous detections.
Technical basis: Mean confidence score across all detected keypoints (joints) for all currently tracked figures. Each keypoint (head, shoulders, elbows, wrists, hips, knees, ankles) is scored independently by the neural network. The average reflects how clearly humanoid the collective detection is.
Technical basis: Inverse-variance of the keypoint confidence distribution across a single detected figure. A real human body produces a detection where most joints score similarly well, the network "sees" all parts of the figure consistently. A false detection caused by environmental features (furniture, clothing, equipment) produces "Frankenstein" detections: a few joints score highly while most score near zero. Kinematic Coherence penalizes this inconsistency.
A score above 0.7 means the figure's joint confidence distribution is internally consistent, it looks like a real body from a detection geometry standpoint, not noise assembling itself into a false positive.
Detection Archive & Evidence Management
The Detection Archive is SPECTER's autonomous evidence management system. Every detection event, at any class level, is logged without investigator intervention. The archive runs independently of the investigator's attention state. Anomalies that occur while you are reacting to a different event, setting up equipment, or reviewing a previous capture are still logged in full.
5.1 Gallery View
Press G or tap GALLERY to open the master session archive. Gallery view displays all session data cards sorted chronologically. Each card contains:
| ELEMENT | DESCRIPTION |
|---|---|
| Snapshot Thumbnail | Full-resolution frame capture at the moment of detection event. Preserves both RGB and depth channel data where available. |
| Detection Type | Which subsystem triggered the event: SPECTRAL ANOMALY, ENTITY DETECTION, FIELD DISTURBANCE, or COMPOSITE (multiple systems triggered simultaneously). |
| Intensity Class | CLASS-I, CLASS-II, or CLASS-III at time of detection. |
| Confidence % | Anomaly Index score at the moment of event fire. |
| Depth Badge | A cyan "D" badge indicates the capture successfully paired visual data with a complete spatial depth map. Depth-badged captures contain full 3D geometry reviewable in post-investigation analysis. |
| Timestamp | Session-relative and wall-clock timestamp for cross-referencing with external evidence (EVP recordings, EMF logs, observer notes). |
5.2 Evidence Export
All archive entries are stored in a structured format on the local drive, accessible after the investigation session closes. Depth-badged captures export with embedded point cloud data. Archive packages are self-contained and suitable for sharing with other investigators, archiving to external media, or uploading to team evidence management systems.
Remote Monitor Application
The SPECTER Remote Monitor is a companion application that connects to a running SPECTER session over a local network, delivering a live feed of system metrics and detection events to a second device, a tablet, a secondary laptop, or a team member's phone positioned elsewhere in the investigation environment.
6.1 Use Cases
Multi-room investigations: Run SPECTER on a fixed rig in a primary investigation room while a second operator monitors remotely from an adjacent room or command post. Detection events are pushed to the remote display in real time, full metric readouts, detection class, and confidence score, without the second operator needing physical access to the primary rig.
Evidence review during active investigation: A second investigator monitors the archive feed live while the primary investigator focuses on the physical environment. When a CLASS-II or CLASS-III event fires, the monitoring operator can begin post-analysis while the primary investigation continues uninterrupted.
Team-based investigations: With the Remote Monitor app running on each team member's device, the entire team shares a synchronized live view of system state regardless of physical location within the site. SPECTER becomes a shared awareness system rather than a single-operator instrument.
6.2 Technical Architecture
The Remote Monitor connects via the SPECTER local server, which runs on the same machine as the primary application. No internet connection is required, the connection is entirely local-network based, preserving operational security in remote investigation sites with no cellular or broadband access. The server is activated from the SPECTER settings panel and assigns a local IP and port displayed on screen. The Remote Monitor app connects to this address.
The remote feed transmits compressed live frame data, all six system metric values, active detector states, and detection event notifications. Full-resolution archive captures are accessible via the remote interface but are stored and managed on the primary device.
Field Protocols
The following protocols represent operational best practices developed through field testing across a range of investigation environments. Adherence to these procedures maximises evidence quality and minimises false positive generation.
7.1 Session Initialisation
- Connect sensor hardware before launching SPECTER. The system will auto-detect the connected sensor on startup and confirm in the source label.
- Allow the session to run for a minimum of 30 seconds before beginning active investigation. This allows the dynamic baseline to establish on the current environment. The log will display BASELINE NOMINAL when the system has calibrated sufficiently.
- Select the appropriate ENV preset for the investigation medium: NEUTRAL for ambient air, FOG for smoke machine investigations, LOW LIGHT for near-darkness operations, OPEN AIR for exterior investigations.
- Set detection class based on expected environment activity. CLASS-II (60%) is recommended for general investigation. Drop to CLASS-I (40%) in low-activity environments. Raise to CLASS-III (80%) in environments with high ambient motion (HVAC, wind, multiple investigators).
7.2 Sector Transitions
- When moving the sensor to a new physical room or sector, allow six seconds of static positioning before resuming active sweep.
- Verify the log displays BASELINE NOMINAL before beginning the sector sweep.
- New sector transitions generate a brief spike in Baseline Deviation as the system recalibrates, this is expected and not anomalous.
7.3 Event Response Protocol
- On a CLASS-I event: Note timestamp. Continue sweep. The system has logged the event automatically.
- On a CLASS-II event: Pause active sweep. Hold sensor position. Allow the system to continue monitoring. Review the flagged region visually. The archive entry is already created, focus on environmental confirmation, not evidence capture.
- On a CLASS-III event: Full stop. Hold all movement in the investigation space. Allow 30 seconds of continued monitoring at the event location. The system will continue logging. Note any corroborating evidence from other instruments (EMF, temperature, audio).
7.4 Sensor Accuracy & Evidence Interpretation
SPECTER mathematically processes exactly what its physical sensors record. The system does not introduce artifacts, extrapolate beyond sensor data, or generate synthetic readings. Every number displayed on screen is derived from a direct sensor input at the hardware level.
Investigators are strongly advised against dismissing elevated readings as system errors without first ruling out environmental variables. The system's detection architecture is conservative by design, it requires multi-metric corroboration before firing a high-confidence event. A CLASS-III detection represents a convergence of anomalies across multiple independent sensor channels simultaneously.
Hardware Setup & System Requirements
8.1 System Requirements
| COMPONENT | MINIMUM | RECOMMENDED |
|---|---|---|
| Operating System | Windows 10 x64 (version 1903+) | Windows 11 x64 |
| Processor | Intel Core i5 / AMD Ryzen 5 (4-core) | Intel Core i7 / AMD Ryzen 7 (8-core) |
| RAM | 8GB | 16GB |
| USB | USB 3.0 Type-A or Type-C | USB 3.1 Gen 2 or Thunderbolt 3 |
| Storage | 10GB free (SSD) | 50GB+ free SSD (for evidence archive) |
| Display | 1080p | 1440p or higher |
8.2 Intel RealSense D435i Setup
- Connect the RealSense D435i to a USB 3.0 or higher port. USB 2.0 is not supported and will produce reduced frame rates.
- SPECTER includes all required RealSense SDK components. No separate driver installation is required.
- On first connection, allow 10–15 seconds for sensor initialisation. The IMU requires a brief calibration period before telemetry readings stabilise.
- For field deployment, the RealSense is powered entirely by USB, no external power supply is required. Estimated operational duration on a fully charged laptop: 4–8 hours depending on hardware.
8.3 Microsoft Kinect V2 Setup
- Install the Kinect for Windows Runtime before launching SPECTER. Available from Microsoft directly. SPECTER will display a sensor error on startup if the runtime is absent.
- The Kinect V2 requires a USB 3.0 port and a dedicated power supply (included with the sensor).
- Kinect V2 is still widely available through secondary markets at approximately $60–$100 USD. New units may also be sourced through specialist electronics retailers.
- IMU telemetry is not available on the Kinect V2. The Field Disturbance Vector metric will display as inactive in the left rail.
8.4 HDMI Capture Card Setup
- Connect your HDMI capture card to a USB 3.0 port. USB-C capture cards are supported.
- Connect your camera (Sony FX3, night vision unit, FLIR thermal imager, or any HDMI-out device) to the capture card's HDMI input.
- In SPECTER settings, select HDMI CAPTURE as the active source. The system will list available capture devices.
- Ensure your camera is outputting a clean HDMI signal (no overlay/menu data in the frame). Most cameras include a "clean HDMI out" setting in the display menu.
- In HDMI capture mode, depth sensor features (SLS Mode, RGB-D capture, IMU telemetry) are not available. Anomaly Mode operates on the full RGB frame from the external camera.
- Recommended capture cards: Elgato Cam Link 4K, AVerMedia Live Gamer Portable, or equivalent USB capture devices with UVC (USB Video Class) compatibility.