Entrophy code mode switching for frequency-domain audio coding

A method of decoding a sequence of encoded audio frequency coefficients with two or more different entropy encoders, the method comprising:

• receiving a coded audio input sequence having a one-dimensional frequency domain range partitioned into at least first and second frequency domain sub-ranges for entropy decoding with different entropy decoders, each of the first and second frequency domain sub-ranges having an associated entropy decoder, wherein range partitions are predetermined, and wherein selection of an appropriate associated decoder is automatic according to the frequency domain sub-range of the coefficients being decoded;
• decoding coefficients of the first frequency domain sub-range with a first associated entropy decoder;
• and
• decoding coefficients of the second frequency domain sub-range with a second associated entropy decoder.

A method of entropy encoding a sequence of audio data frequency coefficients with at least two different entropy encoders, such coefficients having a minimum and a maximum amplitude, the method comprising:

• preparing a first code book for a first entropy encoder;
• preparing a second code book for a second entropy encoder optimized for encoding repeating coefficients of a single value;
• partitioning a one-dimensional frequency domain range for a sequence of audio data frequency coefficients into at least first and second frequency domain sub-ranges so as to facilitate entropy encoding coefficients of different frequency domain sub-ranges with different entropy encoders, such partitioning made according to a statistical analysis identifying separate frequency domain sub-ranges based upon results of coding with the first and second entropy encoders;
• encoding coefficients of the first frequency domain sub-range with the first entropy encoder;
• and
• encoding coefficients of the second frequency domain sub-range with the second entropy encoder.

A system for encoding an input signal with multiple encoding methods according to characteristics of the input signal, the system comprising:

• means for receiving a time-domain audio input signal;
• means for converting the time-domain audio signal to a frequency-domain audio signal;
• means for quantizing the frequency-domain signal into quantized symbols;
• and
• means for selecting an entropy encoder for the quantized symbols based upon spectral frequency of the quantized symbols, wherein the means for selecting associates one of plural different entropy encoders with a spectral frequency depending upon a probability that quantized symbols of the spectral frequency share a value, said associations initially set to minimize output bit-rate of encoding with the different entropy encoders.

A computer readable medium having stored thereon computer-executable instructions for performing a method of encoding audio data with an entropy encoder, the entropy encoder adaptively switching entropy coding modes between frequency domain sub-ranges, the method comprising:

• setting one or more entropy coding mode boundaries in a one-dimensional frequency domain range, wherein each entropy code mode boundary partitions the frequency domain range into plural frequency domain sub-ranges for encoding with different entropy encoding modes, each entropy coding mode boundary for separating audio data that satisfies a statistical criterion from audio data that fails the statistical criterion, the one or more entropy coding mode boundaries initially set to minimize output bit-rate of encoding with the different entropy encoding modes;
• based upon the one or more entropy coding mode boundaries, selecting entropy coding modes to apply to audio data;
• and
• encoding the audio data with the selected entropy coding modes.

A method of encoding a sequence of time-domain audio data symbols with two or more different entropy encoders into an output data stream, each encoder optimized for data sharing a particular value characteristic, the method comprising:

• converting the time-domain data symbols into frequency domain data;
• reducing the frequency domain data according to a psychoacoustic model in conjunction with quantizing the frequency domain data;
• based upon a frequency domain partition, selecting an entropy encoder to apply to the quantized data, the frequency domain partition selectively set to group frequency domain zones that share a particular value characteristic and to separate frequency domain zones for encoding with different entropy encoders, a first group of frequency domain zones characterized by non-zero-value quantized data, a second group of frequency domain zones characterized by zero-value quantized data, the frequency domain partition set based upon results of coding with the different entropy encoders;
• and
• encoding the quantized data with the selected entropy encoder.

A method of encoding an arbitrarily long series of audio input coefficients within a one-dimensional frequency domain range, the method comprising:

• calculating the probability of coefficients having zero values;
• partitioning the one-dimensional frequency domain range into a first frequency domain sub-range and a second frequency domain sub-range to group coefficients having primarily non-zero values into the first frequency domain sub-range, and to group coefficients having primarily zero values into the second frequency domain sub-range;
• encoding coefficients of the first frequency domain sub-range with a variable-to-variable entropy encoder;
• and
• encoding coefficients of the second frequency domain sub-range with a run-length entropy encoder.

A system for encoding an input signal with multiple encoding methods according to characteristics of the input signal, the system comprising:

• an input for receiving a time-domain audio input signal;
• a signal transformer for converting the time-domain audio signal to a frequency-domain audio signal;
• a quantizer for converting the frequency-domain audio signal into quantized symbols;
• and
• a mode selector for selecting an entropy encoder for the quantized symbols depending upon one-dimensional frequency domain zones of the quantized symbols, wherein the mode selector adjustably associates a different entropy encoder with each of plural one-dimensional frequency domain zones based upon a value characteristic of the quantized symbols of the one-dimensional frequency domain zone, and said associations between the plural one-dimensional frequency domain zones and the different entropy encoders facilitate decoder-selection without encoder-designating flags.