Deep Neural Network for FM Broadcast Audio Fidelity
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Terrestrial radio broadcast signals, particularly FM broadcasts, are limited by a maximum audio frequency range of 15 kHz, which restricts the transmission of audio content above this frequency, resulting in incomplete audio reproduction for listeners with human hearing ranges extending beyond 20 kHz.
Innovation Solution
A deep neural network (DNN) module is trained on synthesized audio data and source material audio to reconstruct audio content above the broadcast cutoff frequency, using a synthesized broadcasting processor that adds signal impairments and processes signals through a transmit and receive module, including modulation and demodulation, to enhance FM broadcast signals and restore missing frequencies.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If broadcast signals use conventional bandwidth limitations (15 kHz cutoff for FM), then signal transmission is simplified and compatible with existing standards, but audio fidelity and frequency range are degraded
Solution Approach 1:
The DNN model is trained in advance on pairs of high-resolution audio and bandwidth-limited audio to learn the mapping relationship. This preliminary training enables the model to reconstruct high-frequency content during actual broadcast reception without requiring complex real-time processing or modification of the broadcast signal transmission process.
2Loss of information
If broadcast signals transmit only up to 15 kHz frequency, then bandwidth usage is reduced and transmission is more efficient, but audio content completeness is lost for frequencies above cutoff
Solution Approach 1:
The DNN module acts as an intermediary between the bandwidth-limited broadcast signal and the listener's audio output. It reconstructs missing high-frequency content by learning from training data, effectively mediating between the constrained input signal and the desired high-fidelity output without requiring increased broadcast bandwidth.
3Adaptability or versatility
If receiver modules output digital audio signal with broadcast cutoff frequency limitation, then compliance with broadcasting standards is maintained, but human hearing range coverage is insufficient
Solution Approach 1:
The audio processing is segmented into two distinct stages: (1) reception of standard-compliant bandwidth-limited broadcast signals, and (2) post-reception enhancement using DNN to extend frequency coverage. This segmentation allows the system to maintain standard compliance at the reception stage while achieving extended hearing range coverage at the output stage.
Data Source
AI summary
An apparatus and method for enhancing broadcast radio includes a DNN trained on data sets of audio created from a synthesized broadcasting process and original source audio. Broadcast radio signals are received at a radio module and processed through the DNN to produce enhanced audio.