Real-Time Signal Processing Using Delayed Control Parameters
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Signal processing techniques, particularly in real-time digital signal processing, face significant latency issues due to buffering requirements, which are problematic in applications like live music where high latencies are audible and convolutional neural networks introduce unacceptable delays, prohibiting real-time processing.
Innovation Solution
A signal processing method that generates control data portions with a non-zero latency period, allowing output data to be generated using the most recently available control data portion rather than waiting for control data associated with the same time period, thereby reducing latency and enabling real-time processing.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If control data is generated by processing sampled source signal with Fourier transform or convolutional neural networks, then transformation accuracy and adaptability are improved, but latency increases to unacceptable levels (9ms to 500ms or more)
Solution Approach 1:
The patent applies preliminary action by generating control data for a future time period (second time period) before the current processing time, based on source signal portions from an earlier time period. This allows the output to be generated without waiting for control data to be processed in real-time, thereby reducing latency while maintaining transformation accuracy through pre-computed control parameters.
2Reliability
If Fourier transform with large analysis window is used for spectral domain transformation, then transformation robustness is improved, but buffering latency increases to 4096 samples or 9 milliseconds
Solution Approach 1:
The system performs preliminary computation of control data using a large Fourier analysis window on historical source signal portions, storing these control parameters for future use. When processing current source signal portions, the pre-computed control data from the earlier time period is reused, eliminating the need to wait for current control data computation and thus reducing buffering latency while maintaining robust transformation.
3Adaptability or versatility
If convolutional neural networks are used for signal processing, then processing adaptability is improved, but latency increases to 500 milliseconds or more which prohibits real-time processing
Solution Approach 1:
The patent applies preliminary action by using convolutional neural networks to generate control data in advance for a future time period based on source signal portions from an earlier time period. These pre-computed control parameters are then reused for processing current source signal portions, enabling real-time output generation without the 500ms or more latency that would otherwise prohibit real-time processing while maintaining the adaptability benefits of neural network processing.
Data Source
AI summary
A signal processing method, especially for audio signals, comprises: receiving input signal samples having source signal portions associated with respective time periods (In_n−1); generating parameters associated with the respective time periods by processing the respective input signal samples, in which the step of generating the parameters has an associated non-zero latency period so that generating step provides a parameter p in a time period (T_n−1) later than the start of the respective time period; and generating an output data portion associated with a given time period (Out_n−1 using a signal processing function F which depends upon the input signal samples in the given time period (In_n−1) and an approximated parameter p′ (1110) associated with a time period other than the given time period (T_n−2).