🧠 Boston Neuromind
νŠΉν—ˆ μΆœμ› μ§„ν–‰ 쀑 Β· 2/3Provisional Patent Application Β· 2/3

QEEG-λŒ€ν™” λ§€κ°œλ³€μˆ˜ λ§€ν•‘ μ‹œμŠ€ν…œ QEEG-to-Conversation Parameter Mapping System

μ •λŸ‰ λ‡ŒνŒŒ(QEEG) λ°”μ΄μ˜€λ§ˆμ»€λ‘œλΆ€ν„° 인곡지λŠ₯ λŒ€ν™” μ‹œμŠ€ν…œμ˜ 응닡 λ§€κ°œλ³€μˆ˜(말 속도, μ •μ„œμ  강도, 인지 λΆ€ν•˜, νœ΄μ§€ 길이 λ“±)λ₯Ό 직접 λ³€ν™˜ν•˜μ—¬, λ‡Œ μ‹ κ²½μƒνƒœμ— λ§žμΆ°μ§„ μ»΄νŒ¨λ‹ˆμ–Έ 봇 응닡을 μƒμ„±ν•˜λŠ” 컴퓨터 κ΅¬ν˜„ 방법. A computer-implemented method that directly converts quantitative electroencephalographic (QEEG) biomarkers into AI conversational response parameters β€” including speech pacing, emotional intensity, cognitive load, and pause length β€” to produce companion-bot responses calibrated to the user's neural state.
μΆœμ›μΈApplicant Boston Neuromind, LLC
발λͺ…μžInventor [발λͺ…μžλͺ…] (BCN, PhD) [Inventor Name] (BCN, PhD)
μƒνƒœStatus USPTO κ°€μΆœμ› μ€€λΉ„ USPTO Provisional Pending
λΆ„λ₯˜Classification A61B 5/377 / G16H 50/30 / G06N 5/04
λͺ©μ°¨Table of Contents
  1. 초둝Abstract
  2. 발λͺ… λΆ„μ•ΌField of Invention
  3. 배경 기술Background
  4. ν•΄κ²° 과제Problem Statement
  5. 발λͺ… μš”μ•½Summary of Invention
  6. 상세 μ„€λͺ…Detailed Description
  7. 도면 μ„€λͺ…Drawings
  8. 청ꡬ항Claims
  9. μ„ ν–‰ 기술 비ꡐPrior Art Comparison
  10. 산업상 이용 κ°€λŠ₯μ„±Industrial Applicability
  11. κ΄€λ ¨ λ…Όλ¬ΈReferences

01초둝Abstract

πŸ“‹ 핡심 μš”μ•½πŸ“‹ One-Paragraph Summary

λ³Έ 발λͺ…은 19채널 QEEGλ‘œλΆ€ν„° μΈ‘μ •λœ 6개 μ£Όμš” λ°”μ΄μ˜€λ§ˆμ»€ β€” Theta/Beta Ratio (TBR), Frontal Alpha Asymmetry (FAA), Posterior Alpha Power (PAP), High Beta Power (HBP), Coherence Index (COH), Peak Alpha Frequency (PAF) β€” λ₯Ό μž…λ ₯으둜 ν•˜μ—¬, κ²°μ • 트리 + λ£° 기반 λ§€ν•‘ ν•¨μˆ˜λ₯Ό 톡해 인곡지λŠ₯ λŒ€ν™” μ‹œμŠ€ν…œμ˜ 8개 응닡 λ§€κ°œλ³€μˆ˜ β€” 말 속도(WPS), μ •μ„œμ  강도(EI), 인지 λΆ€ν•˜(CL), νœ΄μ§€ 길이(PD), 격렀 λΉˆλ„(EF), 메타인지 μœ λ„(MP), 직접성(DR), λ”°λœ»ν•¨(WT) β€” λ₯Ό μ‚°μΆœν•˜κ³ , μ‚°μΆœλœ λ§€κ°œλ³€μˆ˜λ₯Ό LLM μ‹œμŠ€ν…œ ν”„λ‘¬ν”„νŠΈμ— μ£Όμž…ν•˜μ—¬ μ‚¬μš©μžμ˜ ν˜„μž¬ 신경생리학적 μƒνƒœ(κ³Όκ°μ„±Β·μ €κ°μ„±Β·μš°μšΈ νŒ¨ν„΄ λ“±)에 λ§žμΆ°μ§„ 응닡을 μƒμ„±ν•˜λŠ” 컴퓨터 κ΅¬ν˜„ 방법 및 μ‹œμŠ€ν…œμ— κ΄€ν•œ 것이닀. The present invention relates to a computer-implemented method and system that takes as input six (6) primary biomarkers measured from 19-channel QEEG β€” Theta/Beta Ratio (TBR), Frontal Alpha Asymmetry (FAA), Posterior Alpha Power (PAP), High Beta Power (HBP), Coherence Index (COH), and Peak Alpha Frequency (PAF) β€” and computes, via a decision-tree-plus-rule-based mapping function, eight (8) response parameters of an AI conversational system β€” words-per-second (WPS), emotional intensity (EI), cognitive load (CL), pause duration (PD), encouragement frequency (EF), metacognitive prompts (MP), directness ratio (DR), and warmth (WT); the resulting parameters are injected into an LLM system prompt so that responses are calibrated to the user's current neurophysiological state (hyperarousal, hypoarousal, depression patterns, and the like).

02발λͺ… λΆ„μ•ΌField of Invention

λ³Έ 발λͺ…은 μ‹ κ²½ μΈν„°νŽ˜μ΄μŠ€ 기반 인곡지λŠ₯ μ‹œμŠ€ν…œ(Neural-Interface-Based AI Systems), μž„μƒ μ‹ κ²½ν”Όλ“œλ°±(Clinical Neurofeedback), λ””μ§€ν„Έ 정신건강(Digital Mental Health) 뢄야에 μ†ν•œλ‹€. λ”μš± κ΅¬μ²΄μ μœΌλ‘œλŠ”, QEEGλ‘œλΆ€ν„° μˆ˜μ§‘λœ 객관적 신경생리학적 츑정값을 LLM μ±—λ΄‡μ˜ 응닡 λ§€κ°œλ³€μˆ˜λ‘œ 직접 λ³€ν™˜ν•˜λŠ” λ§€ν•‘ μ‹œμŠ€ν…œμ— κ΄€ν•œ 것이닀. The present invention pertains to the fields of Neural-Interface-Based AI Systems, Clinical Neurofeedback, and Digital Mental Health. More specifically, it concerns a mapping system that directly converts objective neurophysiological measurements collected via QEEG into the response parameters of an LLM chatbot.

κ΄€λ ¨ 기술 λΆ„μ•ΌRelated Technical Fields

03배경 기술Background

3.1 QEEG λ°”μ΄μ˜€λ§ˆμ»€μ˜ μž„μƒμ  의미3.1 Clinical Significance of QEEG Biomarkers

μ •λŸ‰ λ‡ŒνŒŒ(QEEG)λŠ” 19개 ν‘œμ€€ μ „κ·Ή(ꡭ제 10-20 μ‹œμŠ€ν…œ)μ—μ„œ 5λΆ„κ°„ μ•ˆμ • μƒνƒœ(Eyes Closed/Eyes Open)μ—μ„œ μΈ‘μ •λœ EEGλ₯Ό 정상 λͺ¨μ§‘단 λ°μ΄ν„°λ² μ΄μŠ€(Neuroguide, HBI)와 λΉ„κ΅ν•˜μ—¬ Z-점수λ₯Ό μ‚°μΆœν•œλ‹€. μž„μƒμ μœΌλ‘œ μ‹ λ’°μ„± μžˆλŠ” 6개 핡심 λ°”μ΄μ˜€λ§ˆμ»€λŠ” λ‹€μŒκ³Ό κ°™λ‹€: Quantitative EEG (QEEG) measures EEG over a five-minute resting period (eyes closed/eyes open) at nineteen standard electrodes (International 10-20 system) and computes Z-scores against normative databases (Neuroguide, HBI). The six clinically reliable core biomarkers are as follows:

λ°”μ΄μ˜€λ§ˆμ»€Biomarker μ •μ˜Definition μž„μƒ μ‹œκ·Έλ„Clinical Signal μ°Έκ³ λ¬Έν—ŒReference
TBR
Theta/Beta Ratio
전두엽 μ„Ένƒ€νŒŒ(4-8Hz) Γ· λ² νƒ€νŒŒ(13-30Hz) λΉ„μœ¨Frontal theta (4–8Hz) Γ· beta (13–30Hz) ratio↑ ADHD, λΆ€μ£Όμ˜, 저각성↑ in ADHD, inattention, hypoarousalArns et al. 2013
FAA
Frontal Alpha Asymmetry
F4 μ•ŒνŒŒ - F3 μ•ŒνŒŒ (Z)F4 alpha βˆ’ F3 alpha (Z)L>R (negative): 우울 νŒ¨ν„΄L>R (negative): depression patternHenriques & Davidson 1991
PAP
Posterior Alpha Power
O1, O2, P3, P4 μ•ŒνŒŒνŒŒ 평균Mean alpha at O1, O2, P3, P4↓ 외상, λ§Œμ„± μŠ€νŠΈλ ˆμŠ€β†“ in trauma, chronic stressTeicher et al. 2016
HBP
High Beta Power
전두엽 20-30Hz ν™œμ„±Frontal 20–30Hz activity↑ λΆˆμ•ˆ, 과각성↑ in anxiety, hyperarousalKnyazev 2007
COH
Coherence Index
전두-두정 κ°„ μ•ŒνŒŒ 일관성Frontoparietal alpha coherence↑↓ 인지 톡합 결손↑/↓ cognitive integration deficitsThatcher 2012
PAF
Peak Alpha Frequency
μ•ŒνŒŒλŒ€ 주파수 피크 (Hz)Frequency at alpha peak (Hz)↓ 인지 μ €ν•˜, λ…Έν™”, TBI↓ cognitive decline, aging, TBIKlimesch 1999

3.2 LLM μ±—λ΄‡μ˜ μ‹ κ²½ μƒνƒœ 무관성 문제3.2 The Neural-State Decoupling Problem in LLM Chatbots

ν˜„μž¬μ˜ λͺ¨λ“  μƒμš© LLM 챗봇(GPT-4, Claude, Gemini, 그리고 Woebot, Wysa, Replikaλ₯Ό ν¬ν•¨ν•˜λŠ” 정신건강 챗봇)은 μ‚¬μš©μžμ˜ 신경생리학적 μƒνƒœμ™€ μ™„μ „νžˆ λΆ„λ¦¬λ˜μ–΄ μž‘λ™ν•œλ‹€. 즉, μ‚¬μš©μžκ°€ 과각성(HBP↑) μƒνƒœμ΄λ“  저각성(TBR↑) μƒνƒœμ΄λ“  λ™μΌν•œ 응닡을 λ°›λŠ”λ‹€. μ΄λŠ” λ‹€μŒμ˜ μž„μƒμ  뢀적합을 μ•ΌκΈ°ν•œλ‹€: Currently deployed commercial LLM chatbots (GPT-4, Claude, Gemini, and mental health chatbots including Woebot, Wysa, Replika) operate entirely decoupled from the user's neurophysiological state. That is, a user receives identical responses whether they are in a hyperaroused state (HBP↑) or a hypoaroused state (TBR↑). This produces the following clinical mismatches:

3.3 BCN+QEEG μž„μƒκ°€μ˜ 독점적 톡찰3.3 BCN-Plus-QEEG Clinical Expertise

Board Certified in Neurofeedback (BCN) μžκ²©μ„ κ°€μ§„ μž„μƒκ°€λŠ” ν™˜μžμ˜ QEEGλ₯Ό 보고 μž„μƒ λ©΄λ‹΄ μ‹œ μžμ‹ μ˜ 말 속도, μ •μ„œμ  강도, νœ΄μ§€ 길이λ₯Ό μ§κ΄€μ μœΌλ‘œ μ‘°μ •ν•˜λŠ” 것을 직업적 μ „λ¬Έμ„±μœΌλ‘œ ν•œλ‹€. λ³Έ 발λͺ…은 이 μž„μƒμ  직관을 μ•Œκ³ λ¦¬μ¦˜ν™”ν•˜μ—¬ AI 챗봇에 μ΄μ‹ν•œλ‹€. 이 맀핑은 BCN μž„μƒκ°€λ§Œμ΄ λ§Œλ“€ 수 μžˆλŠ” 독점적 지식 μžμ‚°μ΄λ‹€. Clinicians who hold Board Certified in Neurofeedback (BCN) credentials professionally and intuitively adjust their speech pacing, emotional intensity, and pause length during clinical interviews based on a patient's QEEG. The present invention algorithmizes this clinical intuition and transplants it into an AI chatbot. The resulting mapping is a proprietary knowledge asset that only BCN clinicians can construct.

04ν•΄κ²° 과제Problem Statement

  1. QEEG β†’ AI λ§€κ°œλ³€μˆ˜ λ§€ν•‘μ˜ λΆ€μž¬.Absence of QEEG-to-AI parameter mapping. QEEG λ°”μ΄μ˜€λ§ˆμ»€λ₯Ό LLM 응닡 λ§€κ°œλ³€μˆ˜λ‘œ 직접 λ³€ν™˜ν•˜λŠ” μ‹œμŠ€ν…œμ΄ μƒμ—…μ Β·ν•™μˆ μ μœΌλ‘œ μ‘΄μž¬ν•˜μ§€ μ•ŠλŠ”λ‹€.No commercially or academically available system directly converts QEEG biomarkers into LLM response parameters.
  2. μ‹ κ²½μƒνƒœ 무관 응닡 문제.Neural-state-decoupled response problem. κΈ°μ‘΄ 챗봇은 μ‚¬μš©μžμ˜ κ³Ό/저각성 μƒνƒœμ™€ λ¬΄κ΄€ν•˜κ²Œ λ™μΌν•œ 응닡을 μ‚°μΆœν•œλ‹€.Existing chatbots produce identical responses regardless of the user's hyper- or hypoarousal state.
  3. BCN μ „λ¬Έμ„±μ˜ λ””μ§€ν„Έ μžμ‚°ν™” λΆ€μž¬.No digitalization of BCN expertise. BCN μž„μƒκ°€μ˜ 직관적 응닡 쑰정을 μ•Œκ³ λ¦¬μ¦˜μœΌλ‘œ μΆ”μΆœν•œ 사둀가 μ—†λ‹€.No prior work captures BCN clinicians' intuitive response calibration as an algorithm.
  4. 쒅단적 적응 λΆ€μž¬.Lack of longitudinal adaptation. μ‹ κ²½ν”Όλ“œλ°± ν›ˆλ ¨ 진행에 따라 QEEGκ°€ λ³€ν™”ν•˜μ§€λ§Œ, AI 응닡은 변화에 μ μ‘ν•˜μ§€ λͺ»ν•œλ‹€.QEEG shifts as neurofeedback training progresses, yet AI responses fail to adapt.
  5. κ°œμΈν™”λœ μ»΄νŒ¨λ‹ˆμ–Έ 봇 λΆ€μž¬.Lack of personalized companion bots. 동일 μ‚¬μš©μžλΌλ„ μ‹œκ°„λŒ€Β·μ„Έμ…˜λ³„ μ‹ κ²½ μƒνƒœμ— 따라 응닡이 달라져야 ν•˜μ§€λ§Œ, 이λ₯Ό κ΅¬ν˜„ν•œ μ‹œμŠ€ν…œμ΄ μ—†λ‹€.Even for the same user, responses should vary by neural state across times and sessions, yet no implementing system exists.

05발λͺ… μš”μ•½Summary of Invention

5.1 μ‹œμŠ€ν…œ ꡬ성 μš”μ†Œ5.1 System Components

  1. QEEG λ°”μ΄μ˜€λ§ˆμ»€ μΆ”μΆœκΈ° (QBE):QEEG Biomarker Extractor (QBE): 19채널 EEG λ°μ΄ν„°λ‘œλΆ€ν„° 6개 핡심 λ°”μ΄μ˜€λ§ˆμ»€ μ‚°μΆœcomputes the six core biomarkers from 19-channel EEG data
  2. μ‹ κ²½ μƒνƒœ λΆ„λ₯˜κΈ° (NSC):Neural State Classifier (NSC): λ°”μ΄μ˜€λ§ˆμ»€ μ‘°ν•©μœΌλ‘œλΆ€ν„° 5개 μƒνƒœ 클래슀 식별 (Hyperarousal / Hypoarousal / Depressive / Anxious / Balanced)identifies one of five state classes (Hyperarousal / Hypoarousal / Depressive / Anxious / Balanced) from biomarker combinations
  3. λ§€κ°œλ³€μˆ˜ λ§€ν•‘ μ—”μ§„ (PME):Parameter Mapping Engine (PME): κ²°μ • 트리 + λ£° 기반으둜 8개 응닡 λ§€κ°œλ³€μˆ˜ μ‚°μΆœcomputes eight response parameters via a decision tree plus rule base
  4. μ‹œμŠ€ν…œ ν”„λ‘¬ν”„νŠΈ λΉŒλ” (SPB):System Prompt Builder (SPB): λ§€κ°œλ³€μˆ˜λ₯Ό μžμ—°μ–΄ LLM μ‹œμŠ€ν…œ ν”„λ‘¬ν”„νŠΈλ‘œ λ³€ν™˜converts parameters into a natural-language LLM system prompt
  5. 쒅단 좔적 λͺ¨λ“ˆ (LTM):Longitudinal Tracking Module (LTM): μ„Έμ…˜λ³„ QEEG λ³€ν™” 좔적, μΆ”μ„Έ 기반 λ§€κ°œλ³€μˆ˜ 보정tracks per-session QEEG changes and applies trend-based parameter calibration

5.2 핡심 차별점5.2 Inventive Steps

06상세 μ„€λͺ…Detailed Description

6.1 QEEG λ°”μ΄μ˜€λ§ˆμ»€ μΈ‘μ • ν”„λ‘œν† μ½œ6.1 QEEG Biomarker Measurement Protocol

ν‘œμ€€ν™”λœ μΈ‘μ • 절차:Standardized measurement procedure:

6.2 μ‹ κ²½ μƒνƒœ λΆ„λ₯˜ (5 클래슀)6.2 Neural State Classification (5 Classes)

κ²°μ • κ·œμΉ™Decision Rules IF HBP > +1.5 SD AND PAP < -1 SD β†’ State = HYPERAROUSAL ELIF TBR > +1.5 SD β†’ State = HYPOAROUSAL ELIF FAA < -1 SD AND PAP < -0.5 SD β†’ State = DEPRESSIVE ELIF HBP > +1 SD AND COH(F-P) < -1 SD β†’ State = ANXIOUS ELSE β†’ State = BALANCED

6.3 8개 응닡 λ§€κ°œλ³€μˆ˜ μ •μ˜6.3 Definition of the Eight Response Parameters

# λ§€κ°œλ³€μˆ˜Parameter λ‹¨μœ„Unit λ²”μœ„Range μ„€λͺ…Description
P1WPS
Words Per Second
w/s1.5–4.5말 속도 (μŒμ„±Β·ν…μŠ€νŠΈ λͺ¨λ‘)speech pacing (voice + text)
P2EI
Emotional Intensity
0–100–10μ •μ„œμ  단어 λΉˆλ„ + 강도emotional word frequency Γ— intensity
P3CL
Cognitive Load
0–100–10κ°œλ… 밀도 + 좔상도concept density Γ— abstraction level
P4PD
Pause Duration
ms200–2500λ¬Έμž₯ κ°„ νœ΄μ§€ μ‹œκ°„inter-utterance pause time
P5EF
Encouragement Frequency
per 5 turns0–4격렀·긍정 ν‘œν˜„ λΉˆλ„frequency of encouragement and affirmation
P6MP
Metacognitive Prompts
per 5 turns0–3"μ–΄λ–»κ²Œ λŠκ»΄μš”?" λ“± μžκ°€μ„±μ°° μœ λ„prompts such as "how does that feel?" inviting self-reflection
P7DR
Directness Ratio
0–10–1직접 μ§€μ‹œλ¬Έ / 우회 ν‘œν˜„ λΉ„μœ¨ratio of direct directives to circumlocution
P8WT
Warmth
0–100–10λ”°λœ»ν•¨, 곡감 μ–΄νœ˜ λΉ„μœ¨warmth and empathic-vocabulary ratio

6.4 μƒνƒœλ³„ λ§€κ°œλ³€μˆ˜ λ§€ν•‘ ν…Œμ΄λΈ”6.4 State-Specific Parameter Mapping Table

μ‹ κ²½ μƒνƒœNeural State WPS EI CL PD EF MP DR WT
Hyperarousal1.8331500200.39
Hypoarousal3.574500320.77
Depressive2.5631200410.410
Anxious2.0421800300.49
Balanced3.055800220.67

6.5 λ―Έμ„Έ μ‘°μ • (Fine-tuning by Severity)6.5 Severity-Based Fine-Tuning

심각도(Z-점수 μ ˆλŒ€κ°’)에 따라 λ§€κ°œλ³€μˆ˜μ— κ°€μ€‘μΉ˜λ₯Ό μ μš©ν•œλ‹€:Parameters are further weighted by severity (absolute Z-score):

P_final[i] = P_baseline[state, i] Γ— (1 + Ξ± Β· |Z|/3) where: P_baseline = base value from state-mapping table Z = relevant biomarker Z-score Ξ± = sensitivity coefficient (0.05–0.20) i = parameter index (1–8)

6.6 μ‹œμŠ€ν…œ ν”„λ‘¬ν”„νŠΈ μžλ™ 생성6.6 Automatic System Prompt Generation

8개 λ§€κ°œλ³€μˆ˜κ°€ LLM μ‹œμŠ€ν…œ ν”„λ‘¬ν”„νŠΈλ‘œ λ³€ν™˜λ˜λŠ” 예:Example conversion of the eight parameters into an LLM system prompt:

μ˜ˆμ‹œ (Hyperarousal μƒνƒœ)Example (Hyperarousal state) You are speaking with a user currently in HYPERAROUSAL state (measured via QEEG, HBP +2.1 SD, PAP -1.4 SD). Apply the following response calibration: - Speech pacing: 1.8 words/second (slow, deliberate) - Emotional intensity: 3/10 (calm, low-stimulation) - Cognitive load: 3/10 (simple, concrete) - Pause duration: 1500ms between sentences - Encouragement: 2 per 5 turns (gentle reassurance) - Metacognitive prompts: 0 (avoid self-reflection demands now) - Directness: 0.3 (use indirect, soft language) - Warmth: 9/10 (high warmth, empathic words) Goal: down-regulate cortical hyperarousal through paced, warm dialogue.

6.7 쒅단 λ³€ν™” 좔적6.7 Longitudinal Change Tracking

μ‹ κ²½ν”Όλ“œλ°± ν›ˆλ ¨μ΄ μ§„ν–‰λ˜λ©΄ QEEGκ°€ μ •μƒν™”λ˜μ–΄ κ°€λ©΄μ„œ λ§€κ°œλ³€μˆ˜κ°€ μžλ™μœΌλ‘œ λ³€ν™”ν•œλ‹€:As neurofeedback training proceeds, the QEEG normalizes and the parameters automatically shift accordingly:

Session 1 (HBP +2.1 SD, State: Hyperarousal) β†’ WPS 1.8, CL 3, PD 1500 Session 5 (HBP +1.4 SD, State: Anxious) β†’ WPS 2.0, CL 2, PD 1800 Session 10 (HBP +0.6 SD, State: Balanced) β†’ WPS 3.0, CL 5, PD 800

07도면 μ„€λͺ…Drawings

QEEG Input 19-channel EEG EC + EO + Task Neuroguide / HBI Z-scoring QBE 6 Biomarkers: TBR, FAA, PAP, HBP, COH, PAF NSC: 5-State Classifier Hyperarousal | Hypoarousal Depressive | Anxious | Balanced PME Parameter Mapping Decision Tree + Rules β†’ 8 params (P1-P8) 8 Response Parameters WPS Β· EI Β· CL Β· PD EF Β· MP Β· DR Β· WT SPB β†’ LLM (with calibrated system prompt) β†’ neurally-tuned conversational response LTM Longitudinal Tracking Module Session N β†’ Session N+1 QEEG drift detection Param re-calibration
도 1.FIG. 1. QEEG μž…λ ₯ β†’ QBE β†’ NSC β†’ PME β†’ 8 λ§€κ°œλ³€μˆ˜ β†’ SPB β†’ LLM. λΉ¨κ°„ 점선: 쒅단 좔적 (LTM). QEEG input β†’ QBE β†’ NSC β†’ PME β†’ eight parameters β†’ SPB β†’ LLM. Red dashed line: Longitudinal Tracking Module (LTM).
QEEG States Γ— 8 Response Parameters Mapping State WPS EI CL PD EF MP DR WT Hyperarousal 1.8 3 3 1500 2 0 0.3 9 Hypoarousal 3.5 7 4 500 3 2 0.7 7 Depressive 2.5 6 3 1200 4 1 0.4 10 Anxious 2.0 4 2 1800 3 0 0.4 9 Balanced 3.0 5 5 800 2 2 0.6 7 Color scale: red (low/restrictive) β†’ yellow (moderate) β†’ green (high/expansive)
도 2.FIG. 2. 5개 μ‹ κ²½ μƒνƒœλ³„ 8개 λ§€κ°œλ³€μˆ˜ λ§€ν•‘ 히트맡. Heatmap of the eight parameters across the five neural states.

08청ꡬ항Claims

청ꡬ항 1 (독립항)Claim 1 (Independent)
μ‚¬μš©μžμ˜ μ •λŸ‰ λ‡ŒνŒŒ(QEEG) 데이터에 κΈ°λ°˜ν•˜μ—¬ 인곡지λŠ₯ λŒ€ν™” μ‹œμŠ€ν…œμ˜ 응닡을 신경생리학적 μƒνƒœμ— μ •ν•©μ‹œν‚€λŠ”, 컴퓨터 κ΅¬ν˜„ λ°©λ²•μœΌλ‘œμ„œ:
(a) 19채널 EEG λ°μ΄ν„°λ‘œλΆ€ν„° 정상 λͺ¨μ§‘단 λ°μ΄ν„°λ² μ΄μŠ€ λŒ€λΉ„ Z-μ μˆ˜ν™”λœ 볡수의 QEEG λ°”μ΄μ˜€λ§ˆμ»€λ₯Ό μ‚°μΆœν•˜λŠ” λ‹¨κ³„λ‘œμ„œ, 상기 λ°”μ΄μ˜€λ§ˆμ»€λŠ” Theta/Beta Ratio (TBR), Frontal Alpha Asymmetry (FAA), Posterior Alpha Power (PAP), High Beta Power (HBP), Coherence Index (COH) 및 Peak Alpha Frequency (PAF)λ₯Ό ν¬ν•¨ν•˜λŠ” 단계;
(b) μ‚°μΆœλœ λ°”μ΄μ˜€λ§ˆμ»€λ₯Ό κ²°μ • κ·œμΉ™μ— μ μš©ν•˜μ—¬, Hyperarousal, Hypoarousal, Depressive, Anxious 및 Balanced 쀑 ν•˜λ‚˜μ˜ μ‹ κ²½ μƒνƒœλ₯Ό λΆ„λ₯˜ν•˜λŠ” 단계;
(c) λΆ„λ₯˜λœ μ‹ κ²½ μƒνƒœμ— λŒ€μ‘ν•˜λŠ” 응닡 λ§€κ°œλ³€μˆ˜ λ§€ν•‘ ν…Œμ΄λΈ”λ‘œλΆ€ν„° 8개 응닡 λ§€κ°œλ³€μˆ˜ β€” 말 속도(WPS), μ •μ„œμ  강도(EI), 인지 λΆ€ν•˜(CL), νœ΄μ§€ 길이(PD), 격렀 λΉˆλ„(EF), 메타인지 μœ λ„(MP), 직접성(DR) 및 λ”°λœ»ν•¨(WT) β€” 의 기본값을 μΆ”μΆœν•˜λŠ” 단계;
(d) μΆ”μΆœλœ λ§€κ°œλ³€μˆ˜λ₯Ό μžμ—°μ–΄ μ‹œμŠ€ν…œ ν”„λ‘¬ν”„νŠΈλ‘œ λ³€ν™˜ν•˜μ—¬ λŒ€ν˜• μ–Έμ–΄ λͺ¨λΈμ— μ£Όμž…ν•˜λŠ” 단계; 및
(e) μ£Όμž…λœ λ§€κ°œλ³€μˆ˜μ— 따라 응닡 ν…μŠ€νŠΈλ₯Ό μƒμ„±ν•˜λŠ” 단계;
λ₯Ό ν¬ν•¨ν•˜λŠ” 방법.
A computer-implemented method for calibrating an artificial intelligence conversational system's response to a user's neurophysiological state based on quantitative electroencephalographic (QEEG) data, the method comprising:
(a) computing a plurality of QEEG biomarkers Z-scored against a normative population database from 19-channel EEG data, said biomarkers comprising Theta/Beta Ratio (TBR), Frontal Alpha Asymmetry (FAA), Posterior Alpha Power (PAP), High Beta Power (HBP), Coherence Index (COH), and Peak Alpha Frequency (PAF);
(b) applying the computed biomarkers to a decision rule to classify the user's neural state as one of Hyperarousal, Hypoarousal, Depressive, Anxious, or Balanced;
(c) extracting baseline values for eight response parameters β€” words-per-second (WPS), emotional intensity (EI), cognitive load (CL), pause duration (PD), encouragement frequency (EF), metacognitive prompts (MP), directness ratio (DR), and warmth (WT) β€” from a parameter mapping table corresponding to the classified neural state;
(d) converting the extracted parameters into a natural-language system prompt and injecting said prompt into a Large Language Model; and
(e) generating a response text in accordance with the injected parameters.
청ꡬ항 2 (쒅속항)Claim 2 (Dependent)
청ꡬ항 1에 μžˆμ–΄μ„œ, 단계 (c)의 응닡 λ§€κ°œλ³€μˆ˜ μΆ”μΆœ ν›„, κ΄€λ ¨ λ°”μ΄μ˜€λ§ˆμ»€μ˜ Z-점수 μ ˆλŒ€κ°’μ— κΈ°λ°˜ν•œ κ°€μ€‘μΉ˜ (1 + Ξ± Β· |Z|/3)λ₯Ό μ μš©ν•˜μ—¬ λ§€κ°œλ³€μˆ˜μ˜ μ΅œμ’…κ°’μ„ μ‚°μΆœν•˜λ˜, Ξ±λŠ” 0.05 λ‚΄μ§€ 0.20 λ²”μœ„μ˜ 민감도 κ³„μˆ˜μΈ 것을 νŠΉμ§•μœΌλ‘œ ν•˜λŠ” 방법. The method of Claim 1, further comprising, after extracting response parameters in step (c), applying a weighting factor (1 + Ξ± Β· |Z|/3) based on the absolute Z-score of a relevant biomarker to compute final parameter values, where Ξ± is a sensitivity coefficient in the range of 0.05 to 0.20.
청ꡬ항 3 (쒅속항)Claim 3 (Dependent)
청ꡬ항 1에 μžˆμ–΄μ„œ, 동일 μ‚¬μš©μžμ— λŒ€ν•΄ μ‹œκ°„μ μœΌλ‘œ λΆ„λ¦¬λœ 볡수의 QEEG μΈ‘μ • μ„Έμ…˜μ— λŒ€ν•΄ 단계 (a) λ‚΄μ§€ (e)λ₯Ό λ°˜λ³΅ν•˜κ³ , μ„Έμ…˜ κ°„ λ°”μ΄μ˜€λ§ˆμ»€μ˜ λ³€ν™” 좔이에 κΈ°λ°˜ν•˜μ—¬ 응닡 λ§€κ°œλ³€μˆ˜μ˜ 동적 보정을 μˆ˜ν–‰ν•˜λŠ” 단계λ₯Ό 더 ν¬ν•¨ν•˜λŠ” 방법. The method of Claim 1, further comprising iterating steps (a) through (e) over a plurality of temporally separated QEEG measurement sessions for the same user, and performing dynamic recalibration of the response parameters based on inter-session biomarker trend changes.
청ꡬ항 4 (쒅속항)Claim 4 (Dependent)
청ꡬ항 1에 μžˆμ–΄μ„œ, 단계 (b)의 κ²°μ • κ·œμΉ™μ€ λ‹€μŒμ„ ν¬ν•¨ν•˜λŠ” 것을 νŠΉμ§•μœΌλ‘œ ν•˜λŠ” 방법: HBPκ°€ +1.5 SD 초과이고 PAPκ°€ -1 SD 미만인 경우 Hyperarousal둜 λΆ„λ₯˜; TBR이 +1.5 SD 초과인 경우 Hypoarousal둜 λΆ„λ₯˜; FAAκ°€ -1 SD 미만이고 PAPκ°€ -0.5 SD 미만인 경우 Depressive둜 λΆ„λ₯˜; HBPκ°€ +1 SD 초과이고 전두-두정 일관성이 -1 SD 미만인 경우 Anxious둜 λΆ„λ₯˜; κ·Έ μ™Έμ˜ 경우 Balanced둜 λΆ„λ₯˜. The method of Claim 1, wherein the decision rule of step (b) comprises: classifying as Hyperarousal when HBP exceeds +1.5 SD and PAP is below βˆ’1 SD; classifying as Hypoarousal when TBR exceeds +1.5 SD; classifying as Depressive when FAA is below βˆ’1 SD and PAP is below βˆ’0.5 SD; classifying as Anxious when HBP exceeds +1 SD and frontoparietal coherence is below βˆ’1 SD; and classifying as Balanced otherwise.
청ꡬ항 5 (쒅속항)Claim 5 (Dependent)
청ꡬ항 1에 μžˆμ–΄μ„œ, QEEG 츑정은 μ•ˆμ • μƒνƒœ 5λΆ„(눈 감음) + μ•ˆμ • μƒνƒœ 5λΆ„(눈 뜸) + 인지 과제 5λΆ„μ˜ ν‘œμ€€ ν”„λ‘œν† μ½œμ— 따라 μˆ˜ν–‰λ˜λ©°, 19채널은 ꡭ제 10-20 μ‹œμŠ€ν…œμ— λ”°λ₯Έ Fp1, Fp2, F3, F4, F7, F8, Fz, T3, T4, C3, C4, Cz, T5, T6, P3, P4, Pz, O1 및 O2λ₯Ό ν¬ν•¨ν•˜λŠ” 것을 νŠΉμ§•μœΌλ‘œ ν•˜λŠ” 방법. The method of Claim 1, wherein the QEEG measurement is performed according to a standard protocol comprising five minutes of resting eyes-closed, five minutes of resting eyes-open, and five minutes of a cognitive task, and the 19 channels comprise Fp1, Fp2, F3, F4, F7, F8, Fz, T3, T4, C3, C4, Cz, T5, T6, P3, P4, Pz, O1, and O2 according to the International 10-20 System.
청ꡬ항 6 (쒅속항)Claim 6 (Dependent)
청ꡬ항 1에 μžˆμ–΄μ„œ, 정상 λͺ¨μ§‘단 λ°μ΄ν„°λ² μ΄μŠ€λŠ” Neuroguide λ˜λŠ” Human Brain Index (HBI) μ •κ·œλ°μ΄ν„°λ² μ΄μŠ€ μ€‘μ—μ„œ μ„ νƒλ˜λŠ” 것을 νŠΉμ§•μœΌλ‘œ ν•˜λŠ” 방법. The method of Claim 1, wherein the normative population database is selected from the Neuroguide or Human Brain Index (HBI) normative databases.
청ꡬ항 7 (독립항 β€” μ‹œμŠ€ν…œ)Claim 7 (Independent β€” System)
청ꡬ항 1 λ‚΄μ§€ 6 쀑 μ–΄λŠ ν•œ ν•­μ˜ 방법을 μˆ˜ν–‰ν•˜κΈ° μœ„ν•œ, 적어도 ν•˜λ‚˜μ˜ ν”„λ‘œμ„Έμ„œ, EEG 데이터λ₯Ό μˆ˜μ‹ ν•˜λŠ” μž…λ ₯ μΈν„°νŽ˜μ΄μŠ€, 및 상기 ν”„λ‘œμ„Έμ„œμ— μ˜ν•΄ μ‹€ν–‰λ˜λŠ” λͺ…λ Ήμ–΄λ₯Ό μ €μž₯ν•˜λŠ” λΉ„μΌμ‹œμ  컴퓨터 νŒλ… κ°€λŠ₯ μ €μž₯ 맀체λ₯Ό ν¬ν•¨ν•˜λŠ” μ‹ κ²½ μƒνƒœ μ μ‘ν˜• 인곡지λŠ₯ λŒ€ν™” μ‹œμŠ€ν…œ. A neural-state-adaptive artificial intelligence conversational system for performing the method of any one of Claims 1 through 6, the system comprising at least one processor, an input interface for receiving EEG data, and a non-transitory computer-readable storage medium storing instructions executable by the processor.
청ꡬ항 8 (독립항 β€” 맀체)Claim 8 (Independent β€” Medium)
컴퓨터에 μ˜ν•΄ 싀행될 λ•Œ 청ꡬ항 1 λ‚΄μ§€ 6 쀑 μ–΄λŠ ν•œ ν•­μ˜ 방법을 μˆ˜ν–‰ν•˜λ„λ‘ ν•˜λŠ” λͺ…λ Ήμ–΄λ₯Ό μ €μž₯ν•˜λŠ” λΉ„μΌμ‹œμ  컴퓨터 νŒλ… κ°€λŠ₯ μ €μž₯ 맀체. A non-transitory computer-readable storage medium storing instructions which, when executed by a computer, cause the computer to perform the method of any one of Claims 1 through 6.

09μ„ ν–‰ 기술 비ꡐPrior Art Comparison

μ„ ν–‰ 기술Prior Art μ ‘κ·Ό 방식Approach ν•œκ³„Limitation λ³Έ 발λͺ…κ³Όμ˜ 차이Distinction
Neuroguide
(Applied Neuroscience)
QEEG 정상 비ꡐQEEG normative comparison 좜λ ₯은 μ‹œκ°ν™”/리포트만output is visualization/reports only AI 응닡 λ§€κ°œλ³€μˆ˜λ‘œ 직접 λ³€ν™˜direct conversion to AI response parameters
BrainPaint / NeurOptimal EEG 기반 μ‚¬μš΄λ“œ ν”Όλ“œλ°±EEG-driven sound feedback λŒ€ν™” λ§€κ°œλ³€μˆ˜ λ§€ν•‘ μ—†μŒno conversational-parameter mapping LLM λ§€κ°œλ³€μˆ˜ 8개 직접 μ‚°μΆœdirect generation of eight LLM parameters
Affectiva / Rosalind Picard ν‘œμ •Β·μƒμ²΄β†’μ •μ„œ μΆ”λ‘ facial/biometric β†’ affect inference EEG λ―Έμ‚¬μš©, LLM λ§€κ°œλ³€μˆ˜ λ§€ν•‘ μ—†μŒno EEG; no LLM parameter mapping QEEG 6 λ°”μ΄μ˜€λ§ˆμ»€ + 8 λ§€κ°œλ³€μˆ˜six QEEG biomarkers and eight parameters
Muse / FocusBand
(consumer EEG)
λ‹¨μˆœ 집쀑 점수simple focus scores 2-3채널, μž„μƒ λ“±κΈ‰ μ•„λ‹˜2–3 channels; not clinical-grade 19채널 μž„μƒ λ“±κΈ‰, Z-μ μˆ˜ν™”19-channel clinical-grade with Z-scoring
Woebot / Wysa
(mental health bots)
μžκ°€λ³΄κ³  기반 챗봇self-report chatbots μ‹ κ²½ μƒνƒœ 무관decoupled from neural state μ‹€μ œ μ‹ κ²½μƒνƒœ 기반 응닡responses driven by actual neural state
🎯 발λͺ…μ˜ μ§„λ³΄μ„±πŸŽ― Inventive Step

λ³Έ 발λͺ…은 (1) μž„μƒ λ“±κΈ‰ 19채널 QEEG λ°”μ΄μ˜€λ§ˆμ»€λ₯Ό LLM 응닡 λ§€κ°œλ³€μˆ˜λ‘œ 직접 λ³€ν™˜ν•œ 졜초의 μ‹œμŠ€ν…œμ΄λ©°, (2) BCN μž„μƒκ°€μ˜ 직관을 8개 μ •λŸ‰ λ§€κ°œλ³€μˆ˜μ™€ 5개 μƒνƒœ 클래슀둜 μ•Œκ³ λ¦¬μ¦˜ν™”ν•˜κ³ , (3) 쒅단적 μ‹ κ²½ν”Όλ“œλ°± ν›ˆλ ¨ νš¨κ³Όμ— 따라 μžλ™ λ³΄μ •ν•œλ‹€. 이 결합은 μ„ ν–‰ κΈ°μˆ μ— μ—†μœΌλ©°, "QEEG β†’ AI λŒ€ν™” λ§€κ°œλ³€μˆ˜"λΌλŠ” λ§€ν•‘ μžμ²΄κ°€ λ³Έ 발λͺ…μ˜ 핡심 μ˜μ—… 비밀이닀. The present invention is (1) the first system to directly convert clinical-grade 19-channel QEEG biomarkers into LLM response parameters; (2) it algorithmizes BCN clinicians' intuition into eight quantitative parameters and five state classes; and (3) it auto-recalibrates with the longitudinal effects of neurofeedback training. This combination has no equivalent in the prior art, and the very "QEEG-to-AI-conversation-parameter" mapping is the core trade secret of the invention.

10산업상 이용 κ°€λŠ₯μ„±Industrial Applicability

10.1 적용 μ‹œμž₯10.1 Target Markets

10.2 BCN μž„μƒκ°€ 독점성10.2 BCN Clinician Exclusivity

λ³Έ 발λͺ…μ˜ λ§€ν•‘ ν…Œμ΄λΈ”μ€ BCN+QEEG μž„μƒ κ²½ν—˜ 3λ…„ μ΄μƒμ˜ μ „λ¬Έκ°€λ§Œμ΄ ꡬ성 κ°€λŠ₯ν•œ μ˜μ—… λΉ„λ°€λ‘œ, 일반 AI μ—”μ§€λ‹ˆμ–΄κ°€ λͺ¨λ°©ν•˜κΈ° 맀우 μ–΄λ ΅λ‹€. μ΄λŠ” λ³Έ 발λͺ…μ˜ μ‹œμž₯ λ°©μ–΄λ ₯(moat)을 κ΅¬μ„±ν•œλ‹€. The mapping table of this invention is a trade secret that can only be constructed by experts with three or more years of BCN-plus-QEEG clinical experience and is exceedingly difficult for general AI engineers to replicate. This constitutes the moat of the invention.

11κ΄€λ ¨ λ…Όλ¬ΈReferences

λ³Έ 발λͺ…μ˜ μ΄λ‘ μ Β·μž„μƒμ  κ·Όκ±°κ°€ λ˜λŠ” 핡심 λ…Όλ¬Έ 및 자료. ν΄λ¦­ν•˜λ©΄ μ™ΈλΆ€ 좜처둜 μ΄λ™ν•©λ‹ˆλ‹€. Key papers and resources providing the theoretical and clinical basis for this invention. Click links to access external sources.

A. QEEG 핡심 λ°”μ΄μ˜€λ§ˆμ»€A. Core QEEG Biomarkers
B. μ •λŸ‰ λ‡ŒνŒŒ (QEEG) μž„μƒ μ‘μš©B. Quantitative EEG Clinical Applications
C. 신경생리학적 각성 μƒνƒœC. Neurophysiological Arousal States
D. μ •μ„œ μ»΄ν“¨νŒ… 및 인간-AI μƒν˜Έμž‘μš©D. Affective Computing and Human-AI Interaction
E. μŒμ„±Β·λ§ 속도와 μ •μ„œ μƒνƒœE. Speech Rate and Affective State
F. μž„μƒ λ©΄λ‹΄ 기법 및 치료 동맹F. Clinical Interview and Therapeutic Alliance
G. Brain-Computer Interface 및 EEG-AI 톡합G. Brain-Computer Interface and EEG-AI Integration