Academia.eduAcademia.edu

Figure 10 – uploaded by Bachir Boudraa

See full PDFdownloadDownload figure
Fig. 10. Set partition of an eight-PSK constellation. One important step used in TCM is set partitioning technique [18]. The main goal of this technique is to provide subsets of channel symbols with maximal minimum distance within each subset. Fig. 10 illustrates an example of the set partitioning concept for an eight-PSK signal set. As shown in this figure, the set partitioning divides the overall signal set (channel symbols)

Figure 10 Set partition of an eight-PSK constellation. One important step used in TCM is set partitioning technique [18]. The main goal of this technique is to provide subsets of channel symbols with maximal minimum distance within each subset. Fig. 10 illustrates an example of the set partitioning concept for an eight-PSK signal set. As shown in this figure, the set partitioning divides the overall signal set (channel symbols)

Related Figures (22)

Fig. 3. (a) A section of labelled Ungerboeck’s four-state amplitude modulation trellis. (b) Output alphabet levels and partition. (c) Corresponding TCQ convolutional coder.
Fig. 3. (a) A section of labelled Ungerboeck’s four-state amplitude modulation trellis. (b) Output alphabet levels and partition. (c) Corresponding TCQ convolutional coder.
TCQ performance for the memoryless uniform source using Rate (R + 1) Lloyd-Max output alphabet
TCQ performance for the memoryless uniform source using Rate (R + 1) Lloyd-Max output alphabet
TCQ performance for the memoryless Gaussian source using Rate (R + 1) Lloyd- Max output alphabet Table 3
TCQ performance for the memoryless Gaussian source using Rate (R + 1) Lloyd- Max output alphabet Table 3
TCQ performance for the memoryless Laplacian source using Rate (R + 1) Lloyd-Max output alphabet
TCQ performance for the memoryless Laplacian source using Rate (R + 1) Lloyd-Max output alphabet
Performances of TCVQ encoding with integer rates for the Gaussian source Table 5 Table 6
Performances of TCVQ encoding with integer rates for the Gaussian source Table 5 Table 6
Performances of TCVQ encoding with integer rates for the Laplacian source
Performances of TCVQ encoding with integer rates for the Laplacian source
Performances of TCVQ encoding with integer rates for the Uniform source Table 4
Performances of TCVQ encoding with integer rates for the Uniform source Table 4
SQNR performance of fractional rate TCVQ encoding for the Uniform source Table 7 Table 8
SQNR performance of fractional rate TCVQ encoding for the Uniform source Table 7 Table 8
SQNR performance of fractional rate TCVQ encoding for the Gaussian source
SQNR performance of fractional rate TCVQ encoding for the Gaussian source
SQNR performance of fractional rate TCVQ encoding for the Laplacian source
SQNR performance of fractional rate TCVQ encoding for the Laplacian source
Fig. 5. Comparative evaluation between the TCVQ and its optimized version for Gaussian source encoding. Fig. 4. Comparative evaluation between the TCVQ and its optimized version for Uniform source encoding.
Fig. 5. Comparative evaluation between the TCVQ and its optimized version for Gaussian source encoding. Fig. 4. Comparative evaluation between the TCVQ and its optimized version for Uniform source encoding.
Fig. 6. Comparative evaluation between the TCVQ and its optimized version for Laplacian source encoding.
Fig. 6. Comparative evaluation between the TCVQ and its optimized version for Laplacian source encoding.
Bit allocations of each LSF-TCVQ trellis stage codebook as a function of bit rate Table 14 Table 13
Bit allocations of each LSF-TCVQ trellis stage codebook as a function of bit rate Table 14 Table 13
Average spectral distortion performance of the LSF-TCVQ encoder as a function of bit rate
Average spectral distortion performance of the LSF-TCVQ encoder as a function of bit rate
Average spectral distortion performance of the LSF-TCVQ encoder using the weighted distance measure
Average spectral distortion performance of the LSF-TCVQ encoder using the weighted distance measure
Synthetic speech signal (with 34 bits/frame QS for FS1016's LPCs) Fig. 7. Waveforms and SSNRs for the 3.2 speech sentence ‘“‘Allow leeway here, but rationalize all errors”: (a) Original speech signal, (b) Synthetic speech signal (with LSF-TCVQ for FS1016’s LPCs), (c) Synthetic speech signal (with 34 bits/frame QS for FS1016’s LPCs), (d) SSNR of coded Synthetic speech (b), (e) SSNR of coded Synthetic speech (c).
Synthetic speech signal (with 34 bits/frame QS for FS1016's LPCs) Fig. 7. Waveforms and SSNRs for the 3.2 speech sentence ‘“‘Allow leeway here, but rationalize all errors”: (a) Original speech signal, (b) Synthetic speech signal (with LSF-TCVQ for FS1016’s LPCs), (c) Synthetic speech signal (with 34 bits/frame QS for FS1016’s LPCs), (d) SSNR of coded Synthetic speech (b), (e) SSNR of coded Synthetic speech (c).
Fig. 8. Block diagram of the CELP FS1016 speech coder. for the short term synthesis filter which is used to mode the vocal trac of the speaker. The excita- tion signal for this filter is determined for each sub- frame. It is construc scaled codeve sis-by minimizing a percep first excitation vector is ex ive codebook o The adap ctors se ted by the addition of two ected accord -synthesis search procedure. tual error criterion [32,33]. ing to an analy- This is done by racted from an f 256 codevectors, populated by previous excitation sequences. This codebook is used to mode voiced speec (delay-indice) this codeboo selected from ing 512 pseu diagram of t transmitted parameters for coder model the long term h. For each i, and a gain k. The second do-random co his coder is s period Jq IS icities present in sub-frame, an index determined for vec hown the are the 10 stochastic and or (index i;) is a fixed stochastic codebook contain- devectors. The block in Fig. 8. The chosen FS1016 LSP parameters, the adaptive codebooks indices (i, and
Fig. 8. Block diagram of the CELP FS1016 speech coder. for the short term synthesis filter which is used to mode the vocal trac of the speaker. The excita- tion signal for this filter is determined for each sub- frame. It is construc scaled codeve sis-by minimizing a percep first excitation vector is ex ive codebook o The adap ctors se ted by the addition of two ected accord -synthesis search procedure. tual error criterion [32,33]. ing to an analy- This is done by racted from an f 256 codevectors, populated by previous excitation sequences. This codebook is used to mode voiced speec (delay-indice) this codeboo selected from ing 512 pseu diagram of t transmitted parameters for coder model the long term h. For each i, and a gain k. The second do-random co his coder is s period Jq IS icities present in sub-frame, an index determined for vec hown the are the 10 stochastic and or (index i;) is a fixed stochastic codebook contain- devectors. The block in Fig. 8. The chosen FS1016 LSP parameters, the adaptive codebooks indices (i, and
4800 bps FS1016 bits allocation ig) aS Well as the associated gains (g, and g,).The LSP parameters are quantized using 34 bit scalar quantization, according to the bit pattern (3,4,4,4,4,3,3,3,3,3) and using a non uniform quantization [34]. Table 16 shows the bit alloca- tion scheme for the 4800 bps FS1016 CELP coder parameters transmitted every 30ms through a clear channel.
4800 bps FS1016 bits allocation ig) aS Well as the associated gains (g, and g,).The LSP parameters are quantized using 34 bit scalar quantization, according to the bit pattern (3,4,4,4,4,3,3,3,3,3) and using a non uniform quantization [34]. Table 16 shows the bit alloca- tion scheme for the 4800 bps FS1016 CELP coder parameters transmitted every 30ms through a clear channel.
Fig. 9. General structure of TCM system.
Fig. 9. General structure of TCM system.
Related topics:
EngineeringTechnologySignal ProcessingVector QuantizationLINEAR PREDICTIVE CODINGSpeech CodingScalar Quantization
580 California St., Suite 400
San Francisco, CA, 94104
© 2025 Academia. All rights reserved