Odiosoft.tech AI Based Hearing Enhancement analyzes audiogram data with artificial intelligence to determine the type, severity, and affected frequencies of hearing loss. It then enhances speech intelligibility through a real-time audio processing chain personalized to each patient's audiogram.

The system is an end-to-end hearing assistant that brings clinical audiologic assessment directly to the patient's headphones — no specialized hearing aid hardware required.

Core Capabilities

• AI analysis of audiogram data with audiologist and speech therapist expertise
• Per-frequency correction recommendation (corrected_db + gain_db) for each ear
• Frequency lowering (transposition) for phoneme recovery in high-frequency hearing loss
• Real-time speech enhancement through a Web Audio API processing chain
• Remote speaker support via WebRTC for telehealth scenarios

The system bridges clinical audiology and everyday hearing assistance — analyzing hearing loss, computing corrections, and delivering enhanced audio in real time through standard headphones.

Audiogram Data

• Air conduction thresholds at 8 standard frequencies (250–8000 Hz)
• Bone conduction thresholds for each ear
• Right ear marked in red, left ear in blue (clinical convention)
• Two input methods: AI image capture or manual entry

Hearing Loss Assessment

• Hearing loss type: Conductive / Sensorineural / Mixed
• Severity degree for each ear (mild to profound)
• Overall hearing loss percentage
• Speech discrimination impact evaluation

Affected Frequencies & Phonemes

• Identification of affected frequency ranges
• Affected phonemes: s, ş, f, t, k consonants
• Speech enhancement recommendations per frequency
• Frequency lowering (transposition) recommendations

Per-Frequency Correction

• corrected_db: effective hearing level after amplification (target 20 dB)
• gain_db: amplification needed (original_db − corrected_db, ≥ 0)
• Applied for every frequency where air conduction exceeds 25 dB
• Independent correction for right and left ears

Each frequency is individually evaluated and corrected, ensuring personalized amplification that matches the patient's specific hearing profile.

The system operates in four stages, each using the output of the previous one: Audiogram Input → AI Analysis → Per-Frequency Correction → Real-Time Hearing Enhancement.

Stage 1: Audiogram Input

• AI Image Capture: audiogram photo is taken or uploaded; LLM automatically extracts frequency and dB values
• Manual Entry: air/bone conduction dB values entered manually in the table
• 8 standard frequencies: 250, 500, 1000, 2000, 3000, 4000, 6000, 8000 Hz

Stage 2: AI Audiological Analysis

The AI analyzes audiogram data with audiologist and speech therapist expertise. It determines hearing loss type, severity, affected frequencies, speech discrimination impact, affected phonemes, and computes per-frequency correction for each ear.

Stage 3: Correction Formula

gain_db = original_db − corrected_db (≥ 0)

corrected_db targets 20 dB (normal hearing range). Amplification is applied for every frequency where the air-conduction threshold exceeds 25 dB. The half-gain rule (dB × 0.5) is used as the base amplification strategy.

No Specialized Hardware Required

• Works with standard headphones and any device microphone
• No hearing aid fitting or proprietary hardware needed
• Runs on tablet, laptop, and smartphone
• Results in minutes — repeatable for longitudinal tracking

Advanced Professional Features

Six professional auditory processing features that can be enabled from the advanced settings panel:

• Scene Classification: automatic environment type detection and noise gate adaptation
• Speaker Separation: bandpass 1.7 kHz for speech focus, pushing background speakers back
• Emotional Tone Enhancer: prosody peaking 1.5 kHz boost to preserve emotional intonation
• Phoneme Smart Emphasis: dynamic high-frequency consonant emphasis for clarity
• Latency Optimization: interactive latencyHint for tight lip-sync
• Directional Focus: dual-channel beamforming for nearby speaker focus
• 3-band EQ (Low / Mid / High) with real-time fine-tuning
• Multiband WDRC compression — hearing-aid style loudness compensation
• Remote speaker (WebRTC) support for telehealth
• Automatic audiogram-derived EQ and compression settings

The platform delivers accessible hearing care using existing consumer devices, reducing both cost and environmental impact:

• Accessible screening reduces undiagnosed hearing loss burden, especially in underserved regions
• No consumables or single-use hearing aid batteries — lower environmental footprint
• On-device real-time processing supports data minimization and privacy
• Multilingual access (TR / EN / DE) broadens equitable hearing care
• Works on existing devices — no additional electronic waste from proprietary hardware
• Telehealth-ready: remote speaker support extends care to rural and home-bound patients

By transforming any smartphone, tablet, or laptop into a clinical-grade hearing assistant, the platform democratizes access to audiologic assessment and speech enhancement.

Most hearing enhancement apps apply generic amplification or simple noise reduction. Odiosoft.tech AI Based Hearing Enhancement uniquely combines AI audiogram analysis with a real-time, per-frequency personalized audio processing chain.

Key Innovations

• AI-powered audiogram analysis with LLM — determines hearing loss type, severity, affected phonemes, not just amplification
• Per-frequency correction with frequency lowering (transposition) — recovers high-frequency phonemes that generic amplification cannot
• Real-time Web Audio API processing chain — no proprietary DSP hardware required
• AI image capture of audiograms — eliminates manual data entry errors
• Professional hearing-aid features (WDRC, beamforming, scene classification) available on consumer devices
• Remote speaker support via WebRTC — enables telehealth hearing assistance

Clinical-Grade Processing

The real-time processing chain includes noise suppression AudioWorklet, high-pass/low-pass filters, clarity boost, multiband compression, 3-band EQ, per-ear audiogram-matched chains, and a zero-latency safety limiter — all running in the browser.

Each ear is compensated independently with peaking filters at audiogram frequencies, secondary gain at high frequencies, consonant emphasis, and highshelf boost — delivering personalized hearing care that generic apps cannot match.

Privacy (KVKK / GDPR)

Facial data and voice data are sensitive personal data and are considered 'special category' data under KVKK (Turkey) and GDPR (European Union). Therefore, on-device processing of the data or anonymization before sending to the server is critical. Our system processes image and voice recordings only for analysis purposes; it is essential that biometric data is not shared with third parties and explicit consent is obtained.

• Facial and voice/biometric data — 'special category' under KVKK/GDPR
• On-device processing or anonymization before sending to the server should be preferred
• Explicit consent and data minimization: only necessary data is processed
• Data retention and deletion rights must be communicated to the user

Medical Liability

This system is an AI-supported analysis tool and can provide health/risk estimates to the user; however, it can never make a medical diagnosis (e.g. it should not say 'You have sleep apnea'). AI models can make errors; results are only supplementary information to support a physician's clinical assessment.

• AI-supported analysis — does not replace a definitive diagnosis
• Results are not medical advice; doctor consultation is required
• False positive/negative results are possible; the final decision belongs to the physician
• The decision to start or change treatment must be made only by a specialist physician

Results should be considered as a clinical decision support tool and used as a screening and triage layer that does not replace a physician examination or, when necessary, laboratory tests.