Project Description
Case Studies / AI & ML
AI & ML
Enhancing Patient Experience with Intelligent Age and Gender Detection
Our client, a healthcare technology company, prioritized demographic insights to enhance patient care.
Book a strategy call
The Client
Client Overview
Our client, a healthcare technology company, prioritized demographic insights to enhance patient care. Focusing on telemedicine and remote monitoring, they required an advanced, speech-based age and gender detection system. Seamless collection of demographic data during patient interactions was aimed at enabling personalized care plans, empowering healthcare professionals, and enhancing care quality.
01
Challenges
Build a system that predicts the age and gender of the user based on the user’s speech.
The system should be highly accurate in predicting both age and gender from spoken utterances.
The system should be evaluated for biases that may lead to unfair predictions across different demographic groups.
02
Approach
Technical implementation
- Collected TIMIT, NISP, and noise datasets, covering various accents, languages, and speaking styles.
- Preprocessed the speech data to distribute gender data evenly within the datasets.
- Designed and utilized a multi-scale deep learning architecture with three parallel CNNs, each operating on different kernel sizes.
- Defined and configured the hyperparameters.
- Extracted features through convolutional and pooling layers to automatically learn feature representations, and to detect patterns and structures in the spectrograms.
- Implemented a regression task for predicting age and mapped age ranges to numerical values.
- Applied a classification task for predicting gender and used binary encoding for gender.
- Trained the model using the preprocessed features and labeled data from the TIMIT and NISP datasets on NVIDIA A100 GPU.
- Evaluated the model on a validation set, measuring metrics such as accuracy, mean squared error, and mean absolute error.
03
Impact
Achieved a Mean Squared Error of 5.6 for age detection and an accuracy of 97% for gender detection with demographic variations, languages, and accents.
Demonstrated consistent and unbiased age predictions with less than 6% variation in performance across diverse demographic groups.
Improved patient throughput by 7% and reduced administrative costs by 9% with automated data collection and processing.
Increased telehealth utilization rates by 13% due to enhanced effectiveness and personalized experiences.
Want results like these from your own data?
Talk to a specialistSee how you could get similar results
Rudder Analytics builds data and AI systems that ship — and stay tied to accuracy, cost, and customer experience.
