Peak GRU Neural Network for Hearing Device Audio Processing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing audio processing technologies, such as hearing aids, face challenges in efficiently reducing noise while maintaining low power consumption, especially in applications where data changes are limited over time, like in audio or video processing with small variations between time steps.
Innovation Solution
A modified gated recurrent unit (GRU) neural network, termed Peak GRU, is implemented in hearing devices, which limits processing to only the peak values that have changed significantly, reducing computational load and power consumption by discarding smaller changes.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a neural network processes all audio data channels at every time step, then noise reduction quality is improved, but power consumption increases
Solution Approach 1:
The patent extracts only the most significant changes in audio data by calculating differences between consecutive time steps and selecting only those exceeding a threshold. This extraction approach processes only relevant data portions rather than all data, reducing computational load and power consumption while maintaining noise reduction effectiveness
Solution Approach 2:
The patent applies partial action by processing only a subset of audio data channels at each time step - specifically those showing significant changes. Instead of processing all channels continuously, the system selectively processes only the necessary portions, reducing overall computational effort while preserving quality where needed
2Use of energy by moving object
If computational processing is reduced to save power, then power consumption decreases, but noise reduction effectiveness deteriorates
Solution Approach 1:
The patent implements dynamic processing by adapting the amount of computation based on the actual content of the audio signal. The system dynamically determines which time steps and frequency channels require processing based on change thresholds, making the computational effort match the actual needs of the signal rather than using a fixed processing regime
Solution Approach 2:
The patent changes the processing parameters dynamically by adjusting which data points are processed based on threshold comparisons. The system modifies its processing behavior by selecting different subsets of data to process depending on the signal characteristics, thereby maintaining effectiveness while reducing overall computational load
3Measurement precision
If all frequency channels are processed at every time step, then audio quality is maintained, but computational complexity increases
Solution Approach 1:
The patent segments the audio processing task by dividing frequency channels and time steps into groups that require processing and those that don't. By segmenting the data based on change thresholds, the system processes only relevant segments rather than the entire dataset, reducing computational complexity while maintaining quality in processed segments
Solution Approach 2:
The patent applies partial action by processing only a subset of frequency channels at each time step - specifically those showing significant spectral changes. This selective processing approach reduces the number of operations required while ensuring that channels contributing most to audio quality changes are fully processed
Data Source
AI summary
A hearing device, e.g. a hearing aid or a headset, configured to be worn by a comprises an input unit for providing at least one electric input signal in a time-frequency representation; and a signal processor comprising a target signal estimator for providing an estimate of the target signal; a noise estimator for providing an estimate of the noise; and a gain estimator for providing respective gain values in dependence of said target signal estimate and said noise estimate. The gain estimator comprises a trained neural network, wherein the outputs of the neural network comprise real or complex valued gains, or separate real valued gains and real valued phases. The signal processor is configured—at a given time instance t—to calculate changes Δx(i,t)=x(i,t)−{circumflex over (x)}(i,t−1), and Δh(j,t−1)=h(j,t−1)−ĥ(j,t−2) to an input vector x(t) and to the hidden state vector h(t−1), respectively, from one time instance, t−1, to the next, t, and where {circumflex over (x)}(i,t−1) and ĥ(j,t−2) are estimated values of x(i,t−1) and h(j,t−2), respectively, where indices i, j refers to the ith input neuron and the jth neuron of the hidden state, respectively, where 1≤i≤Nch,x and 1≤j≤Nch,oh, wherein Nch,x and Nch,oh is the number of processing channels of the input vector x and the hidden state vector h, respectively, and wherein the signal processor is further configured to provide that the number of updated channels among said Nch,x and said Nch,oh processing channels of the modified gated recurrent unit for said input vector x(t) and said hidden state vector h(t−1), respectively, at said given time instance t is limited to a number of peak values Np,x, and Np,oh, respectively, where Np,x is smaller than Nch,x, and Np,oh, is smaller than Nch,oh.