Voice Harbor par Nijta

PHI Anonymization

PHI anonymization is a critical process for protecting Protected Health Information from unauthorized access and misuse. It transforms healthcare data so it can no longer identify individuals, even when combined with other available data.

This ensures that sensitive patient information such as names, addresses, medical record numbers, diagnosis details, or biometric identifiers are either removed or altered to preserve privacy.

Vertical Audio Flow Component
Audio Input
Original voice recording
Transcription
Convert to text
PHI Detection
Identify sensitive data
Anonymization
Replace with tones
Synthesis
Generate safe audio
PHI Benefits Component

Benefits of PHI anonymization

Enables secure and compliant storage of sensitive voice and text data
Supports legal and ethical data sharing with third parties
Facilitates AI model training and analytics without compromising patient privacy
Preserves the utility of original data for innovation and research
Enhances patient trust by protecting identity and sensitive health details
Anonymization Example Component

Real-World Anonymization Example

Original Transcription
"Hi, this is Dr. Emily Carter, calling regarding patient Michael Reynolds, date of birth 07/09/1982, who had a follow-up scheduled at Stanford Medical Center for March 15th, 2025. His insurance ID is AETN-4455-2391."
Anonymized Output
"Hi, this is <tone>, calling regarding patient <tone>, date of birth <tone>, who had a follow-up scheduled at <tone> for <tone>. His insurance ID is <tone>."
Identified PHI entities:
<DOCTOR NAME> <PATIENT NAME> <DATE OF BIRTH> <HOSPITAL NAME> <APPOINTMENT DATE> <INSURANCE ID>
Clean Blue Features Grid
Secure Storage
Enables secure and compliant storage of sensitive voice and text data.
AI Training
Facilitates AI model training without compromising patient privacy.
Data Utility
Preserves the utility of original data for innovation and research.
PHI Tabbed Interface

HIPAA-defined PHI Identifiers

1
Name
John Doe, Mary Smith
2
Geographic data (smaller than a state)
123 Main St, Springfield, IL 62704
3
Dates related to individual (excluding year)
Birthdate: 05/14/1982, Discharge Date: 09/22
4
Telephone numbers
(555) 123-4567
5
Fax numbers
(555) 987-6543
6
Email addresses
johndoe@example.com
7
Social Security numbers
123-45-6789
8
Medical record numbers
MRN: 567890123
9
Health plan beneficiary numbers
Medicare ID: A123456789
10
Account numbers
Patient account: 00349876
11
Certificate/license numbers
Driver's license: D1234567
12
Vehicle identifiers and serial numbers
License plate: XYZ-789, VIN: 1HGCM82633A004352
13
Device identifiers and serial numbers
Pacemaker SN: 1029384756
14
Web URLs
www.patientportal.com/johndoe
15
IP address numbers
192.168.1.1
16
Biometric identifiers (incl. voiceprints)
Voiceprint used for patient ID; Fingerprint scan
17
Full-face photos and comparable images
Passport photo, ID card image
18
Any unique identifying code or characteristic
Internal system ID: X5A-KL89-P1

Nijta's Unique Advantage - Voiceprint Protection

Voiceprints are explicitly listed under biometric identifiers and are protected PHI under HIPAA.
They can uniquely identify individuals and must be anonymized before data is reused, shared, or stored.

This is a core feature of our platform , we detect and anonymize voiceprints automatically in audio to ensure full HIPAA compliance.

Glass UI Language Support - White Background

Supported Languages

Spanish
3.0%
WER Score
Italian
4.0%
WER Score
English
4.2%
WER Score
German
4.5%
WER Score
French
7.1%
WER Score

Speech-to-Text

Speech-to-Text or Automatic Speech Recognition (ASR) is the process of converting speech or audio into written text. Our advanced ASR technology is built on the robust foundation of OpenAI's Whisper, known for its exceptional performance in multilingual speech recognition. However, we've significantly enhanced its capabilities with in-house innovations, including the implementation of phonetic time-stamps. These detailed markers provide an extra layer of precision by capturing the timing of specific phonetic elements within the audio, enabling more granular analysis and synchronization.

Our ASR component supports multiple languages and has robust code-switching capabilities, as it effortlessly transcribes audio that blends various languages. Its built-in automatic language detection ensures that users do not have to manually specify the language, streamlining the workflow, while precise time-stamps allow for easy navigation and review of audio content.

Benchmarks for Top 10 Supported Languages

Benchmarks for top 10 supported languages

Rank Language WER (%) on FLEURS
1 Spanish 3.0
2 Italian 4.0
3 English 4.2
4 Portuguese 4.3
5 German 4.5
6 Japanese 5.0
7 Polish 5.6
8 Russian 5.6
9 Dutch 6.1
10 Indonesian 6.4

Speaker Diarization

Speaker diarization is the process of identifying and segmenting individual speakers within an audio recording. It plays a crucial role in scenarios where multiple participants are involved, such as meetings, interviews, or call center conversations. By accurately distinguishing between speakers, diarization helps in creating clear, organized transcripts, improving sentiment analysis, and enhancing audio data analytics. This technology is particularly valuable for compliance, customer experience monitoring, and research purposes, where understanding who said what is essential for accurate analysis and reporting.

We are excited to unveil Monster, our new speaker diarization system, Nijta’s latest innovation in audio segmentation that redefines both accuracy and efficiency. This first release marks a major step forward in speaker diarization, offering precise segmentation, multilingual support, and robust performance on your noisy data. From medical conversations to customer service calls and teleconferences, our advanced model adapts to various acoustic conditions, ensuring high accuracy where other diarization systems fall short.

Features Section
Precise speaker diarization
Seamless handling of overlapping speech
Robust performance across varied acoustic environments
Unlimited number of speakers

Biometric Anonymization

Biometric anonymization is the process of altering or removing unique vocal characteristics to prevent speaker re-identification while preserving the usability of the audio. Since voice carries biometric markers such as pitch, tone, and speech patterns, anonymization techniques ensure that speech data remains valuable for transcription, analytics, and AI training without compromising individual privacy. A pseudo voice with a choice of gender is created from a large random pool of speakers to completely prevent re-identification.

This feature is currently available in English and French. Additional languages could be requested for your particular use case.

Libérez vos données
Commencez à utiliser Nijta.

Anonymisez votre premier enregistrement vocal en quelques minutes.