Method and apparatus for processing discontinuities in digital sound signals caused by pitch control

A method for processing small discontinuities caused by a digital pitch shift operation, said method comprising the steps of:

• multiplying a current input signal by a coefficient "a";
• delaying the input signal;
• multiplying the delayed signal by a coefficient "b";
• and
• adding the products of the first and second multiplications to smooth the small discontinuities caused by the pitch shift operation,
• wherein said coefficients "a" and "b" are changed while maintaining a predetermined relationship, and, for the pitch down operation, the pitch down processed data value is moderately approximated to the delayed data value within the period where the small discontinuity occurs, and for the pitch up operation, the pitch up processed data values are moderately approximated to the current data value within the period where the small discontinuity occurs, to reduce the extent of the data discontinuity, and
• the coefficients "a" and "b" satisfy the equation a+b=1, the value "a" decreases from 1 by a pitch shift degree coefficient K for the pitch down operation, and when reaching 0 or less, the same data as one sampling period before is read out again from the memory which stores digital sound sampling data, while for the pitch up operation, the value "a" increases from 0 by a pitch shift degree coefficient K, and when reaching 1 or more, the data to be read out from the memory at that sampling period is skipped.

A method for processing large discontinuities in digital sound signals, comprising the steps of:

• producing two converted data sequences from one original input data sequence, the timings of the large discontinuity occurrence for the data sequences being offset with respect to each other;
• and
• cross fading said two converted data sequences by overlapping them, one data value of the converted signal in which a discontinuity occurs being increased from its discontinuity occurrence point by a weighting coefficient, while the other is decreased by a weighting coefficient from the same point in time,
• said weighting values for the two data sequences varying in time along an upwardly convex polygonal line.

An apparatus for processing discontinuities in a digital sound signal caused by a pitch shift operation, comprising:

• means for storing the input sound data;
• means for inputting a pitch shift extent and for generating an address, when the pitch shift extent represents pitch down, the same address being generated twice every predetermined cycle of periods defined by the pitch shift extent, and when the pitch shift extent represents pitch up, a certain address being skipped every predetermined cycle of periods defined by the pitch shift extent;
• means for reading out two sound data sequences from the input sound data stored in the storage means, based on the addresses generated by the address generating means;
• first smoothing means for inputting and smoothing one of the two sound data sequences;
• second smoothing means for inputting and smoothing the other sound data sequence;
• means for calculating weighting coefficients, based on the pitch shift extent, for each of the smoothing process and applying them to the first and second smoothing means;
• means for cross fading the two smoothed data sequences;
• and
• means for calculating weighting coefficients, based on the pitch shift extent, for cross fading and applying them to the cross fade processing means.

An apparatus for processing large discontinuities in digital sound signals, comprising:

• means for producing two converted data sequences from one original input data sequence, the timings of the large discontinuity occurrence for the data sequences being offset with respect to each other;
• means for cross fading said two converted data sequences by overlapping them, one data value of the converted signal in which the small discontinuity occurs being increased from its discontinuity occurrence point by a weighting coefficient while the other is decreased by a weighting coefficient from the same point in time;
• and
• means for controlling the cross fade weighting values so that the weighting coefficients for the two data sequences vary in time along an upwardly convex polygonal line.