Pitch Lag Predictor for Speech Packet Loss Concealment
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional packet loss concealment algorithms in speech coding, such as G.729, fail to effectively predict pitch lag parameters during consecutive or burst packet losses, leading to significant impairments in speech quality due to the use of static or incrementally copied pitch lag values, resulting in unpleasant click sounds.
Innovation Solution
A pitch lag predictor that calculates a predicted pitch lag parameter using a summation of previous pitch lag values and their positions, with coefficients determined by minimizing an error equation, allowing for a more dynamic and accurate prediction of pitch lags.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If conventional packet loss algorithms copy or duplicate the previous pitch lag parameter for lost frames, then the implementation is simple, but the speech quality deteriorates due to considerable distance gap between reconstructed and true pitch lags causing click sounds
Solution Approach 1:
The patent transitions from static pitch lag copying to dynamic prediction by modeling pitch lag as a function of frame number and previous pitch lags. The predictor uses the equation P(i) = a + b*i where a and b are coefficients derived from minimum mean square error optimization, allowing the predicted pitch lag to adapt dynamically to the actual pitch track behavior rather than remaining fixed or incrementally copied.
Solution Approach 2:
The patent changes the parameter approach from using constant or linearly incrementing pitch lag values to using predicted values based on statistical modeling. The coefficients a and b are calculated using the formulas a=(3*sum0−sum1)/5 and b=(sum1−2*sum0)/10, where sum0 and sum1 are summations of previous pitch lag parameters, enabling the system to adapt to actual pitch variations during packet loss.
2Device complexity
If conventional algorithms use static or incrementally copied pitch lag values during packet loss, then the computational complexity is low, but the speech quality deteriorates due to perceptible click sounds and large gaps in pitch track
Solution Approach 1:
The patent implements feedback by using the actual pitch lag values from received frames to train the prediction model. The coefficients a and b are calculated based on the minimum mean square error between predicted and actual pitch lags, creating a feedback loop where the model continuously improves its accuracy based on observed pitch behavior during and after packet loss events.
Solution Approach 2:
The patent performs preliminary action by pre-calculating the coefficients a and b using minimum mean square error optimization before actual packet loss occurs. The formulas a=(3*sum0−sum1)/5 and b=(sum1−2*sum0)/10 are derived in advance based on the relationship between frame number and pitch lag, enabling the system to be ready for efficient prediction when packet loss actually occurs.
3Ease of manufacture
If pitch lag parameters are duplicated for consecutive lost frames, then the algorithm is simple to implement, but the reliability of speech quality maintenance deteriorates for consecutive or burst losses
Solution Approach 1:
The patent makes the pitch lag prediction dynamic by using the frame number i and previous pitch lag values to calculate predicted values that adapt to the actual pitch track. This dynamic approach ensures that even during consecutive or burst packet losses, the predicted pitch lags remain reliable and close to the true values, maintaining speech quality without the need for complex algorithms.
Data Source
AI summary
There is provided a pitch lag predictor for use by a speech decoder to generate a predicted pitch lag parameter. The pitch lag predictor comprises a summation calculator configured to generate a first summation based on a plurality of previous pitch lag parameters, and a second summation based on a plurality of previous pitch lag parameters and a position of each of the plurality of previous pitch lag parameters with respect to the predicted pitch lag parameter; a coefficient calculator configured to generate a first coefficient using a first equation based on the first summation and the second summation, and a second coefficient using a second equation based on the first summation and the second summation, wherein the first equation is different than the second equation; and a predictor configured to generate the predicted pitch lag parameter based on the first coefficient and the second coefficient.